American Institutes for Research
1000 Thomas Jefferson Street NW, Washington, DC 20007-3835 | 202.403.5000 | TTY 877.334.3499 | www.air.org
REPUBLIC OF ZAMBIA
MINISTRY OF COMMUNITY DEVELOPMENT, MOTHER
AND CHILD HEALTH
SOCIAL CASH TRANSFER SCHEME
24-Month Impact Report for the Child Grant Programme
September 2013
Principal investigators:
Dr. David Seidenfeld, Dr. Sudhanshu Handa, and Dr. Gelson Tembo
1
Contents
Contributors .................................................................................................................................................. 2
Acknowledgments ......................................................................................................................................... 3
Acronyms ...................................................................................................................................................... 4
Executive Summary ....................................................................................................................................... 5
I. Introduction ............................................................................................................................................. 10
II. Conceptual Framework ........................................................................................................................... 12
III. Study Design ........................................................................................................................................... 14
IV. Attrition.................................................................................................................................................. 16
V. Operational Performance ....................................................................................................................... 19
VI. Consumption Expenditures ................................................................................................................... 22
VII. Poverty and Food Security .................................................................................................................... 28
XIII. Young Child Outcomes ......................................................................................................................... 35
IX. Children Over 5 Years Old ...................................................................................................................... 40
X. Women .................................................................................................................................................... 43
XI. Birth Outcomes ...................................................................................................................................... 46
XII. Economic Impacts ................................................................................................................................. 48
XIII. Discussion and Conclusion ................................................................................................................... 63
Annex 1: Prices in the CGP Evaluation Sample ........................................................................................... 67
Annex 2: Difference-in-Differences Estimation .......................................................................................... 69
Annex 3: Mean Differences at Baseline for Attrition Analysis .................................................................... 71
Annex 4: Additional Results on Economics Impacts ................................................................................... 75
Annex 5: The Local Economy-wide Impact Evaluation Model for the CGP ................................................ 76
Annex 6: Community Profile ....................................................................................................................... 77
2
Contributors
The evaluation of the Child Grant Program is being conducted by American Institutes for Research (AIR)
for the Ministry of Community Development, Mother and Child Health in Zambia, under contract to
UNICEF, with funding from the Cooperating Partners— DfID, and Irish Aid. The Principal Investigators for
the overall evaluation are David Seidenfeld (AIR) and Sudhanshu Handa (University of North Carolina at
Chapel Hill). The Zambia-based Principal Investigator is Gelson Tembo of Palm Associates and the
University of Zambia. The FAO (Principal Investigator Benjamin Davis) was contracted by DFID-UK to
provide the analysis of the economic and spillover effects of the CGP. The overall team leaders of this
report are David Seidenfeld (AIR), Sudhanshu Handa (UNC), and Benjamin Davis (FAO), but many others
made important contributions and are listed below by institutional affiliation and alphabetical order
within institution:
AIR: Cassandra Jessee, Leah Prencipe, Dan Sherman
Palm: Alefa Banda, Liseteli Ndiyoi, Nathan Tembo
FAO: Silvio Daidone, Joshua Dewbre, Mario Gonzalez-Flores
UC-Davis: Ed Taylor, Karen Thome
UNC: Amber Peterman
The suggested citation for this report is:
American Institutes for Research. (2013). Zambia’s Child Grant Program: 24-month impact report.
Washington, DC: Author.
Contact information:
David Seidenfeld Sudhanshu Handa Benjamin Davis
[email protected] [email protected] [email protected]
Gelson Tembo
3
Acknowledgments
This evaluation report has been drafted by American Institutes for Research (AIR) on behalf of the
Ministry of Community Development, Mother and Child Health. AIR recognizes the contributions of
many individuals and organizations without whom it would not have been possible to complete this
study. Our thanks go to the Zambian Ministry of Community Development, Mother and Child Health
(MCDMCH); the Department for International Development (DfID); the United Nations Children Fund
(UNICEF); Irish Aid; and Palm Associates for the opportunity to carry out this study and for the financial
and technical support that they rendered.
Our special thanks go to Dr. Gelson Tembo (Palm Associates) for carrying out the data collection and Mr.
Paul Quarles van Ufford (UNICEF) and Ms. Kelley Toole (DfID) for their technical support during the
design and fieldwork. The value of the logistical support obtained from Mr. Stanfield Michelo, the
Director of Social Welfare at the MCDMCH; the staff in the cash transfer unit at the MCDMCH, Lusaka;
and the district social welfare officers (DSWO) in Shangombo, Kaputa, and Kalabo also cannot be
overemphasized. Everyone at the ministry provided valuable logistical support during data collection in
the three districts, including program background information.
Our acknowledgments would be incomplete without mentioning our team of very able research
assistants in Zambia. Specifically, we acknowledge the input of the team of enumerators and supervisors
from Palm Associates, whose dedication during data collection ensured that the data collected were of
high quality. The highly competent team of data entry personnel at Palm Associates is also greatly
acknowledged.
The patience exercised by the Zambian households, community leaders, and community members
during interviews are also greatly acknowledged. It is our hope that the insights from the information
that they provided will translate into valuable interventions in their communities.
David Seidenfeld, Ph.D.
4
Acronyms
AIR American Institutes for Research
ARI Acute Respiratory Illness
CGP Child Grant Social Cash Transfer Program
CWAC Community Welfare Assistance Committee
DD Differences-in-differences
ECD Early Childhood Development
FANTA Food and Nutrition Technical Assistance Project
FAO Food and Agricultural Organization of the United Nations
HDDS Household Dietary Diversity Score
IYCF Infant and Young Child Feeding
LCMS Living Conditions Monitoring Survey
LEWIE Local Economy-wide Impact Evaluation
MCDMCH Ministry of Community Development, Mother and Child Health (MCDMCH)
MICS Multiple Indicators Cluster Surveys
RCT Randomized Controlled Trial
UNICEF United Nations Children's Fund
ZDHS Zambia Demographic and Health Survey
ZMW Zambian Kwacha
ZOI Zone of Influence
5
Executive Summary
Background
In 2010, the Government of the Republic of Zambia through the Ministry of Community Development,
Mother and Child Health (MCDMCH) began implementing the Child Grant social cash transfer program
(CGP) in three districts: Kalabo, Kaputa, and Shangombo. The CGP targets households with children
under age 5 living in program districts and provides each household with 60 kwacha (ZMW), or roughly
U.S. $12, a month, regardless of household size. Payments are made every other month, and there are
no conditions to receive the money. An impact evaluation was conducted as the program was
implemented to learn its effects on recipients and provide evidence for deciding the future of the
program. American Institutes for Research (AIR) was contracted by UNICEF Zambia in 2010 to design and
implement a randomized controlled trial (RCT) for a 3-year impact evaluation of the program and to
conduct the necessary data collection, analysis, and reporting.1 This report presents findings after 24
months of program implementation, including impacts on expenditures, poverty, food security, children
under age 5, children older than 5, and the economy.
Study Design
We implemented an RCT to estimate program impacts after 2 years. This study includes 2,515
households in 90 Community Welfare Assistance Committees (CWACs) that have been randomly
assigned to treatment or control conditions. As shown in the baseline report, randomization created
equivalent groups. We lost 226 households (9 percent) to attrition 2 years into the study; however, we
maintain equivalent groups and find no differential attrition between treatment and control groups. By
maintaining the RCT design, we can attribute observed differences between treatment and control
groups directly to the CGP. At baseline (2010), we hypothesized about where we expected to find
program effects based on the logic model and ex-ante simulations to predict impacts using the baseline
data. We compare these estimates from baseline with observed impacts 2 years later.
Operational Performance
Overall, we find that the Ministry has successfully implemented the cash transfer program. Beneficiaries
receive the correct amount of money according to schedule, can access the money without any cost and
with relative ease, and do not experience unethical solicitations. Although recipients understand the
eligibility criteria to enter the program, they have some misunderstanding about the conditions required
to remain in the program, with many thinking that they need to spend the money to feed or clothe their
children. The results of this study suggest that perceptions of conditions by the recipients might
influence the impact of the program.
Consumption Expenditures
As predicted at baseline, a majority of the increased spending for CGP recipients goes for food (76
percent), followed by health and hygiene (7 percent), clothing (6 percent), and transportation/
communication (6 percent). In contrast, there is no significant program impact for spending on
1 Palm Associates was contracted by AIR to assist with the baseline data collection.
6
education, domestic items, or alcohol/tobacco overall. However, we do find impacts on education
spending for larger households because they have more children. Among the increased food
expenditures, the largest share goes to cereals (40 percent), followed by meats, which include poultry
and fish (21 percent), and then fats (15 percent) and sugars (11 percent). These impacts on food
expenditures differ when we look at them by household size. In smaller households, the impact of the
CGP on food is concentrated on cereals (where 45 percent of the impact on food is derived) followed by
meat (15 percent), fats (14 percent), and pulses (13 percent). However among larger households, the
impact of the grant on food is driven by meats (32 percent) and then cereals (30 percent). The
conceptual framework suggests that the primary direct impact of the CGP will be on the consumption
spending behavior of recipient households. The other outcomes in this study, such as nutrition,
education, and material needs, are second-round effects in that they are not affected directly by the
cash transfer but require a series of behavioral responses by the household induced by the income
effect of the cash transfer in order to change. Therefore, we expect to see second-round impacts that
coincide with observed spending patterns.
Poverty and Food Security
We find strong impacts for reducing extreme poverty and improving food security. The program reduces
the extreme poverty household rate by 5.4 percentage points; however, the largest program impacts
are found for the poverty gap (10.0 percentage points) and squared poverty gap (10.8 percentage
points), which account for the distribution of individuals below the line rather than whether individuals
moved above the line. We also find that the CGP increases the percentage of households eating two or
more meals per day by 8 percentage points, with almost everyone eating two or more meals per day (97
percent). The program increases the number of households that are not severely food insecure by 18
percentage points, a 113 percent improvement over the control group. The CGP has a large impact on
perceptions of food security. Twice as many CGP households (71 percent) as control households (35
percent) do not consider themselves very poor, a 31 percentage point difference. Five times more CGP
households than control households report being better off now than they were 12 months ago, a 45
percentage point increase. These findings are consistent with predicted program impacts at baseline.
Young Children
We find strong impacts on reducing the incidence of diarrhea (4.9 percentage points) for children under
5 years old, but none for other young child health outcomes. The result for diarrhea is consistent with
our hypotheses from baseline. We find no impacts on curative or preventative health-seeking behavior,
which is also consistent with our hypotheses.
We find a large impact of the CGP on infant and young child feeding (IYCF)—an increase of 22
percentage points (from 32 percent to 60 percent, the control group improved to only 43 percent), an
88 percent increase over the baseline mean. This result is consistent with the consumption expenditure
effects as well as the ex-ante predictions. The program also significantly increases weight for height
among children ages 3 to 5. However, we do not find any impacts on height, which corroborates our
baseline hypotheses, which predicted impacts to weight but not to height.
7
Older Children
We find large impacts on material well-being, with a 33 percentage point increase to the number of
children who have all three needs met (shoes, second set of clothing, and a blanket), but no overall
impacts on education or health. These results are supported by the spending patterns observed 24
months into the program and are consistent with baseline predictions. However, we find that the
program has impacts on education outcomes, such as enrollment and attendance, for children with less
educated mothers. This result occurs because more educated mothers have already enrolled their child
and have less room from growth.
Productive Impacts
We find large impacts on crop and livestock production. The CGP increases the amount of operated land
by 18 percentage points (a 34 percent increase from baseline), as well as the use of agricultural inputs.
The program increases the share of households with any expenditure on inputs by 18 percentage points,
from a baseline share of 23 percent. This increase is particularly relevant for smaller households (22
percentage points) and includes spending on seeds, fertilizer, and hired labor.
The increase in crop input use and tool ownership leads to an increase in the value of aggregate
production. The CGP increases the overall value of the harvest by ZMW 146, which is a 50 percent
increase from baseline. The program increases the share of households producing maize by 8
percentage points and 4 percentage points in the share producing rice. The overall increase in
production appears to be destined for sale rather than consumed on farm. The CGP increases the share
of households selling crops by 12 percentage point (an over 50 percent increase from baseline), and we
do not find any increase in the share of consumption out of own production.
The program increases the production of livestock. The CGP has a positive impact on the ownership of a
wide variety of animals, both in terms of share of households with livestock (a 21 percentage point
increase overall, from 49 percent at baseline) and in the total number of different types of poultry.
Further, beneficiary households experience approximately double the volume of purchase and sales of
livestock compared with control households.
Impact to Labor
We find impacts to non-farm business activity and shifts in the labor supply from working on other
people’s farms to focusing on own farm and non-farm enterprises. The share of beneficiary households
operating a non-agricultural enterprise increases by 17 percentage points compared with control
households. Moreover, the program also increases the number of months in operation, the value of
total monthly revenue and profit, and the share of households owning business assets.
The impact of the CGP on the economic activities of beneficiary households implies changes in labor
supply. Overall, we find a significant shift from agricultural wage labor to family agricultural and non-
agricultural businesses, which corresponds with the increases in household level economic activities
brought on by receipt of the CGP transfer. The CGP decreases the share of households with an adult
engaged in wage labor by 9 percentage points, an impact that is stronger for females of working age.
8
Most of the decrease occurs in agricultural wage labor (14 fewer days overall) and is offset by large
significant increases in non-farm work (20 days overall) and non-farm enterprise (1.6 days). The program
does not have any impact on child work for pay.
Local Economy
The CGP is likely to have significant multiplier effects on the local economy. A simulation model shows
that the CGP has a potential total income multiplier of ZMW 1.79. That is, each kwacha transferred to
poor households can raise income in the local economy by ZMW 1.79. Beneficiary households receive
the direct benefit of the transfer, whereas ineligible households receive the bulk of the indirect benefit.
Of the ZMW 1.79 income multiplier, ineligible households would receive ZMW 0.62 for each kwacha
given to beneficiary households, while the beneficiary households receive the value of the transfer plus
an extra ZMW 0.17, for a total of ZMW 1.17 for these recipient households. Beneficiary households thus
benefit both directly and indirectly from the transfer program. More important, though, the CGP also
confers significant benefits to non-beneficiaries through the increased demand for goods and services
generated by their increased purchasing power.
Overall Summary
The CGP has generated positive impacts on a range of indicators identified in the conceptual framework
as being plausible. What is particularly exciting about the results presented here is that the CGP not only
addresses the immediate consumption and food security needs of recipients but also leads to significant
increases in the productive capacity of households, both by supporting the expansion of existing
economic activity by enabling their diversification into new activity. There is also evidence that the
program is beginning to have an impact on young children through improved feeding and reductions in
wasting, as well as older children. The table below links each program objective with the indicators
reported here.
Summary of Impacts in Areas Directly Linked to CGP Objectives
Supplement and not replace household income
Increase of ZMW 15 in monthly per capita consumption
expenditure
Reduction of 11 percentage points in poverty gap and
squared poverty gap
Increase the number of households having a second
meal per day
Increase of 8 percentage points in households with 2+
meals per day
Increase of 22 percentage points in proportion of
children ages 6 to 24 months receiving minimum
feeding requirements
Reduce the rate of mortality and morbidity of children
under 5
Reduction in diarrhea of 5 percentage points
Reduce stunting and wasting among children under 5
Increase in weight-for-height of 0.196 z-scores among
children ages 3 to 5 years
Increase in weight-for-weight and weight-for-age of
0.118 and 0.128, respectively, among children ages 0
to 5, but no statistically significant effects
9
Increase the number of children enrolled in and
attending primary school
No statistically significant effects
Increase the number of households owning assets such
as livestock
Increase of 21 percentage points in households owning
any livestock.
Increase of 4.5 percentage points in households owning
any non-farm business assets.
10
I. Introduction
This paper provides the 24-month follow-up results for the Child Grant cash transfer program impact
evaluation. In 2010, the government of the Republic of Zambia through the Ministry of Community
Development, Mother and Child Health (MCDMCH) began implementing the Child Grant cash transfer
program (CGP) in three districts: Kaputa, Kalabo, and Shongombo. American Institutes for Research (AIR)
was contracted by UNICEF Zambia in 2010 to design and implement a randomized controlled trial (RCT)
for a 3-year impact evaluation of the program and to conduct the necessary data collection, analysis,
and reporting.2 This paper presents findings from the 24-month follow-up study in 13 sections:
Introduction, Conceptual Framework, Study Design, Attrition, Operational Performance, Consumption
Expenditures, Poverty and Food Security, Young Child Outcomes, Children Over 5 Years Old, Women,
Birth Outcomes, Economic Impacts, and Discussion and Conclusion.
Background
In 2010, Zambia’s MCDMCH started the rollout of the CGP in three districts: Kalabo, Kaputa, and
Shongombo. Zambia had been implementing cash transfer programs since 2004 in 12 other districts,
trying different targeting models including community-based targeting, proxy means testing, and
categorical targeting by age (over 60 years old). The government decided to introduce a new model, the
CGP, in three new districts that had never received any cash transfer program. This categorical model
targets any household with a child under 5 years old. Recipient households receive 60 kwacha (ZMW) a
month (equivalent to U.S. $12), an amount deemed sufficient by the MCDMCH to purchase one meal a
day for everyone in the household for 1 month. The amount is the same regardless of household size.
Payments are made every other month through a local pay point manager, and there are no conditions
to receive the money.
Locations
The MCDMCH chose to start the CGP in three districts within Zambia that have the highest rates of
extreme poverty and mortality among children under age 5, thus introducing an element of
geographical targeting to the program. The three districts are Kaputa, located in Northern Province;
Shongombo, located in Western Province; and Kalabo, also located in Western Province. All three
districts are near the Zambian border with either the Democratic Republic of Congo (Kaputa) or Angola
(Shongombo and Kalabo) and require a minimum of 2 days of travel by car to reach from the capital,
Lusaka. Because Shongombo and Kalabo are cut off from Lusaka by a flood plain that turns into a river in
the rainy season, they can be reached only by boat during some months of the year. These districts
represent some of the most remote locations in Zambia, making them a challenge for providing social
services, and are some of the most underprivileged communities in Zambia.
Enrollment
Only households with children under age 3 were enrolled in the program to ensure that every recipient
household receives the transfers for at least 2 years. This means that the baseline sample includes only
2 Palm Associates was contracted by AIR to assist with the baseline data collection.
11
households with a child under 3. The Ministry implements a continuous enrollment system in which
households are immediately enrolled after having a newborn baby. Thus, every household in the district
with a child under 5 will receive benefits for 2 years after the program is introduced to that area.
Objectives
According to the MCDMCH, the goal of the CGP is to reduce extreme poverty and the intergenerational
transfer of poverty. The objectives of the program relate to five primary areas: income, education,
health, food security, and livelihoods. Therefore, the impact evaluation will primarily focus on assessing
change in these areas. The objectives of the program according to the CGP operations manual follow (in
no specific order):
• Supplement and not replace household income
• Increase the number of children enrolled in and attending primary school
• Reduce the rate of mortality and morbidity of children under 5
• Reduce stunting and wasting among children under 5
• Increase the number of households having a second meal per day
• Increase the number of households owning assets such as livestock
12
II. Conceptual Framework
The CGP provides an unconditional cash transfer to households with a child under age 5. CGP-eligible
households are extremely poor, with 95 percent falling below the national extreme poverty line and
having a median household per-capita daily consumption of ZMW 1.05, or approximately 20 U.S. cents.
Among households at such low levels of consumption, the marginal propensity to consume will be
almost 100 percent; that is, they will spend all of any additional income rather than save it. Thus, we
expect the immediate impact of the program will be to raise spending levels, particularly basic spending
needs for food, clothing, and shelter, some of which will influence children’s health, nutrition, and
material well-being. Once immediate basic needs are met, and possibly after a period of time, the
sustained influx of new cash may then trigger further responses within the household economy, for
example, by providing room for investment and other productive activity, the use of services, and the
ability to free up older children from work to attend school.
Figure 2.1 brings together these ideas into a conceptual framework that shows how the CGP can affect
household activity, the causal pathways involved, and the potential moderator and mediator factors.
The diagram is read from left to right. We expect a direct effect of the cash transfer on household
consumption (food security, material well-being), on the use of services, and possibly even on
productive activity after some time. Sociological and economic theories of human behavior suggest that
the impact of the cash may work through several mechanisms (mediators), including a woman’s
bargaining power within the household (because the woman receives the cash directly) and the degree
to which the woman receiving the cash is forward looking. Similarly, the impact of the cash transfer may
be weaker or stronger depending on local conditions in the community. These moderators include
access to markets and other services, prices of goods and services, and shocks. Moderating effects are
shown with dotted lines that intersect with the solid lines to indicate that they can influence the
strength of the direct effect.3
The next step in the causal chain is the effect on children, which we separate into effects on older and
younger children because of the program’s focus on very young children and because the key indicators
of welfare are different for the two age groups. It is important to recognize that any potential impact of
the program on children must work through the household by its effect on spending or time allocation
decisions (including use of services). The link between the household and children can also be
moderated by environmental factors, such as distance to schools or health facilities, as indicated in the
diagram, and household-level characteristics themselves, such as the mother’s literacy. Indeed, from a
theoretical perspective, some factors cited as mediators may actually be moderators, such as women’s
bargaining power. We can test for moderation versus mediation through established statistical
3 A mediator is a factor that can be influenced by the program and so lies directly within the causal chain. A
moderator, in contrast, is not influenced by the program. Thus, service availability is a moderator, whereas
women’s bargaining power may be either a moderator or a mediator depending on whether it is itself changed by
the program. Maternal literacy is a moderator and not a program outcome, unless the program inspires caregivers
to learn to read and write.
13
techniques,4 and this information will be important to help us understand the actual impact of the
program on behavior.
Figure 2.1 identifies some of the key indicators along the causal chain that we analyze in the evaluation
of the CGP. These are consistent with the log frame of the project and are all measured using
established items in existing national sample surveys such as the Living Conditions Monitoring Survey
(LCMS) and the Zambia Demographic and Health Survey (ZDHS). The only exception is the school
readiness indicator, which is a relatively new index developed by UNICEF to be rolled out as part of its
global Multiple Indicators Cluster Surveys (MICS) Program.
Beyond the household: local economy effects
Figure 2.1 provides a framework for understanding the impact of the program on beneficiaries, but
economic theory, and indeed common sense, tells us that significant injections of cash into a small
geographical area can have spillover effects on non-beneficiaries as well. This is because the increased
purchasing power of beneficiaries raises demand for goods and services, which in turn can increase
profits of local businesses if they are able to respond to demand. These local economy, or spillover,
effects, to the extent that they exist, are important to document in order to understand the full impact
of the program on the residents of a beneficiary community.
4
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research:
Conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology, 51(6), 1173–1182.
14
III. Study Design
The CGP impact evaluation relies on a design in which communities were randomized to treatment and
control to estimate the effects of the program on recipients. Communities designated by Community
Welfare Assistance Committees (CWACs) were randomly assigned to either the treatment condition to
start the program in December 2010 or to the control condition. This study reports on the effects of the
program after 2 years.
Benefits of Randomization
A randomized controlled trial (RCT) is the most powerful research design for drawing conclusions about
the impacts of an intervention on specific outcomes. An RCT draws from a pool of comparable subjects
and then randomly assigns some to a treatment group that receives the intervention and others to a
control group that receives the intervention against which comparisons can be made. An RCT permits us
to directly attribute any observed differences between the treatment and control groups to the
intervention; otherwise, other unobserved factors, such as motivation, could have influenced members
of a group to move into a treatment or control group.5 Randomization helps ensure that both observed
and unobserved characteristics that may affect the outcomes are similar between the treatment and
control conditions of the sample. In a randomized experiment, treatment and control groups are
expected to be comparable (with possible chance variation between groups) so that the average
differences in outcome between the two groups at the end of the study can be attributed to the
intervention. Our analysis of comparison and treatment groups finds that randomization created
equivalent groups at baseline for the CGP evaluation (see the baseline report for a complete description
of the randomization process and results).
Timing and Process of Data Collection
To ensure high-quality and valid data, we paid special attention to the process and timing of data
collection, making sure that it was culturally appropriate, sensitive to Zambia’s economic cycle, and
consistently implemented. AIR contracted with Palm Associates, a Zambian research firm with years of
experience conducting household surveys throughout Zambia, to help implement the CGP survey and
enter the data. A team of Zambian enumerators experienced in household and community surveys and
fluent in the local language where they worked were trained on the CGP instrument and then tested in
the field before moving into their assigned communities for data collection.
One enumerator collected data in each household, interviewing the identified potential female recipient
and documenting her answers. This oral interview process was necessary because many of the
recipients are illiterate. In addition to interviewing the female head of household, the enumerator
collected anthropometric measures (height and weight) for every child age 7 or under, using high-quality
height boards and scales endorsed by UNICEF. Enumerators were trained in proper anthropometric
measuring techniques and then supervised in the field by specialists from Zambia’s National Food and
Nutrition Commission. In addition to the household survey, two senior enumerators administered a
5 Campbell, D. T., & Stanley, J. C. (1963). Experimental and quasi-experimental designs for research. Hopewell, NJ:
Houghton Mifflin.
15
community questionnaire in every CWAC to a group of community leaders, including CWAC committee
members, teachers, village headmen, and local business owners. Last, a senior member on the
enumerator team administered four business enterprise questionnaires for each CWAC.
The 24-month follow-up data collection occurred in Zambia’s lean season, when people have the least
amount of food left from the previous harvest and hunger is at its greatest. The timing of this round of
data collection fell exactly 24 months from the baseline study, ensuring that households are being
compared in the same season as at baseline. Furthermore, Zambia’s seasonality was taken into account
to ensure accessibility to households. Zambia has three seasons: a rainy season from December through
March, a cold dry season from April through August, and a hot dry season from September through
November. Data collection was timed early in the lean season, September through October of 2012, to
prevent difficulty reaching households due to flooding. Crops are planted in the rainy season and
harvested throughout the rainy season and into May. Food is most scarce toward the end of the hot dry
season (October and November) because this is the longest period without a food harvest. The CGP aims
to support poor households during this period of hunger by providing enough money to purchase a meal
a day. We believe that the biggest impacts of the program are likely to be observed during this lean
season; thus, the study is designed with baseline and follow-up periods of data collection during this
season.
Data Entry
Palm Associates entered the data as they came in from the field. Data were verified using double entry
on separate computers, flagging inconsistent responses between the two entries, and referring to the
original questionnaire to see the actual response.
Analysis Approach
This study is a longitudinal, randomized, controlled evaluation with repeated measures at the individual
and household levels. We estimate program impacts on individuals and households using a differences-
in-differences (DD) statistical model that compares change in outcomes between baseline and follow-up
and between treatment and control groups (see Annex 2 for details on this method).6 The DD estimator
is the most commonly used estimation technique for impacts of cash transfer models and has been
used, for example, in Mexico’s Progresa program7 and Kenya’s Cash Transfer for Orphans and
Vulnerable Children.8 We use cluster-robust standard errors to account for the lack of independence
across observations due to clustering of households within CWACs.9 We also use inverse probability
weights to account for the 9 percent attrition in the follow-up sample.10 The CGP provides the same
transfer size to a household, regardless of the household size. Therefore, we investigate differential
impacts by household size for each outcome. We present impacts by household size only when they are
6 Local economy effects use a different analysis approach, which is explained in the appendix.
7 http://wbro.oxfordjournals.org/cgi/reprint/20/1/29
8 Kenya CT-OVC Evaluation Team. (2012). The impact of the Kenya CT-OVC Program on human capital. Journal of
Development Effectiveness, 4(1), 38–49. 9 http://www2.sas.com/proceedings/sugi23/Posters/p205.pdf
10 Woolridge, J. W. (2010). Econometric analysis of cross section and panel data. Cambridge, MA: MIT Press.
16
different. Additionally, an influx of cash into a region may influence non-beneficiary households as well,
a phenomenon that is estimated through a local economy model called the (LEWIE) method (see Annex
5).
IV. Attrition
Attrition within a sample occurs when households from the baseline sample are missing in the follow-up
sample. Mobility, the dissolution of households, death, and divorce can cause attrition and make it
difficult to locate a household for a second data collection. Attrition causes problems in conducting an
evaluation because it not only decreases the sample size (leading to less precise estimates of program
impact) but also introduces selection bias to the sample, which will lead to incorrect program impact
estimates or change the characteristics of the sample and affect its generalizability.11 There are two
types of attrition: differential and overall. Differential attrition occurs when the treatment and control
samples differ in the types of individuals who leave the sample. Differential attrition can create biased
samples by eliminating the balance between the treatment and control groups achieved through
randomization at baseline. Overall attrition is the total share of observations missing at follow-up from
the original sample. Overall attrition can change the characteristics of the remaining sample and affect
the ability of the study’s findings to be generalized to populations outside the study. Ideally, both types
should be small.
We investigate attrition at the 24-month follow-up by testing for similarities at baseline between (1)
treatment and control groups for all nonmissing households (differential attrition) and (2) all households
at baseline and the remaining households at the 24-month follow-up (overall attrition). Testing these
groups on baseline characteristics can assess whether the benefits of randomization are preserved at
follow-up. Fortunately, we do not find any significant differential attrition at the 24-month follow-up,
meaning that we preserve the benefits of randomization. We find small differences between the study
population at baseline and those that remain at the 24-month follow-up; the remaining households are
less likely to have experienced a shock, especially flooding or drought at baseline, and they consume a
higher proportion of maize over cassava. The differences from overall attrition are primarily driven by
the lower response rate in Kaputa district.
Differential Attrition
We find no difference in baseline characteristics between the treatment and control households that
remain in the study at the 24-month follow-up, meaning that there is no differential attrition and the
benefits of randomization are preserved. Table 4.1 shows the household response rates at the 24-month
follow-up by treatment status for each district. The response rates are balanced between the treatment
and control groups. We test all the household, young child, and older child outcome measures and
control variables for statistical differences at baseline between the treatment and control groups that
remain in the 24-month follow-up analysis. None of the 43 indicators is statistically different,
demonstrating that on average, people missing from the 24-month follow-up sample looked the same at
baseline regardless of whether they were from the treatment or control group. The similarity of the
11
What Works Clearinghouse (http://ies.ed.gov/ncee/wwc/documentsum.aspx?sid=19)
17
characteristics of people missing in the follow-up sample between treatment statuses allays the concern
that attrition introduced selection bias. Thus, the study maintains strong internal validity created
through randomization, enabling estimated impacts to be attributed to the cash transfer program rather
than to differences in the groups resulting from attrition. See Annex 3 for the results of the tests mean
differences on the 43 indicators.
District Treatment Control n
Kaputa 82.3 80.1 837
Kalabo 96.4 95.9 838
Shangombo 96.4 96.7 839
Overall 91.9 90.6 2514
Table 4.1: Household Response Rate by Study Arm at 24-Month
Follow-Up for CGP (n = 2515)
Overall Attrition
Ninety-one percent of the households from baseline remain in the 24-month follow-up sample. Table
4.2 indicates that 72 percent of the missing households come from Kaputa. Most of the attrition in
Kaputa occurred because the Cheshi lake is drying up, forcing households that relied on the lake for
fishing and farming at baseline to move their homes as they follow the edge of the lake inward. Entire
villages disbanded, with households spreading out to new areas and building new homes in remote
swampy areas that are difficult to locate or reach by vehicle on land. This problem in Kaputa affected
treatment and control households equally, demonstrated by the lack of differential attrition by
treatment status.
District Response
rate
Households at
Baseline
Percent of Total
Missing Households
Kaputa 81 837 72
Kalabo 96 838 15
Shangombo 97 839 13
Overall 91 2514 100
Table 4.2: Overall Attrition for CGP 24-Month Follow-Up:
Household Response Rate by District
There is almost no difference in baseline characteristics between the remaining sample at the 24-month
follow-up and the sample at baseline, with no mean differences on all but two indicators. The relatively
large attrition in Kaputa leads to two small differences in the characteristics of the total sample that
remains at the 24-month follow-up compared with the entire sample at baseline. We find that when
compared with baseline, the remaining sample contains a lower rate of households that experienced
shocks and a lower share of roots and tubers, on average, is consumed. See Table 4.3 for details on
these variables with statistical differences at baseline between the missing households and those that
remain. The larger attrition rate in Kaputa drives these findings because households in Kaputa tend to
eat more cassava (a tuber) as their staple food instead of maize, which is more common in Kalabo and
Shangombo. This cultural difference explains the decrease in the average household consumption share
18
of roots and tubers. The ecological changes in Kaputa region, especially the lake drying up, explain why
we find that a slightly smaller percentage of remaining households reported experiencing a shock such
as drought or flooding (16 percent) compared with the entire baseline sample (19 percent). See Annex 3
for all results comparing the baseline sample with those who remain in the 24-month follow-up.
Table 4.3: Differences Between the Full Sample and the Sample Remaining at the 24-Month Follow-Up
Variables Full
Sample N1
Remaining
Sample N2
Mean
Difference p-value
Roots/tubers share 0.17 2519 0.15 2295 0.02 <.0001
Household affected by any shocks 0.19 2519 0.16 2298 0.03 <.0001
T-tests clustered on the CWAC level.
The remaining sample at 24-month-follow-up is likely more similar to populations throughout Zambia
because most of the missing households from the study depend on a lake that is drying up for their
livelihood, a characteristic less common throughout the country. The ability to generalize results from
the study to populations outside the study area, say, to other districts in Zambia or to other countries,
changes as the study sample that remains changes from baseline. Therefore, the study’s generalizability
(external validity) likely has increased with the new study population that remains at the 24-month
follow-up because the remaining sample is more similar to the populations where the program might be
scaled to.
19
V. Operational Performance
The MCDMCH had been implementing the CGP cash transfer program for 2 years by the time AIR
conducted the follow-up round of data collection. We use this opportunity to investigate the fidelity of
program implementation from the beneficiaries’ perspective. This section discusses the results of the
implementation questions. We focus on two primary areas: payments and program understanding.
Overall, the Ministry successfully implements the cash transfer program. Beneficiaries receive the
designated amount on schedule; they can access the money without any cost and with relative ease;
and they do not experience unethical solicitations. Although recipients understand the eligibility criteria
to enter the program, they have some misunderstandings about the conditions required to remain in
the program, with many thinking that they need to spend the money to feed their children. The analyses
for this section include only responses from beneficiaries of the program at the 2-year follow-up. Thus,
all the data presented here are from people who have been receiving the cash transfers for 2 years.
Data and analyses are presented through descriptive statistics due to the cross-sectional nature of the
data. The 1,128 households in the sample are spread across 45 CWACs in the three CGP districts
(Kaputa, Kalabo, and Shangombo).
Payments
Monitoring payments provides insights into program efficiency. Ineffective payment distribution may
result in underutilization of funds, missed payments, and dissatisfaction in beneficiary households. High
private costs for the recipients, such as expenses to access payment, solicitations or mistreatment by
program staff, and lack of timely payments could have a negative impact on program effects. The
potential problems in distribution could also add upfront costs to the Ministry, making program
expansion within Zambia challenging. This study investigates recipient experiences around four themes
related to payments: access to payments, notifications of payments, unjust solicitations for payments,
and timeliness.
Access: Findings from the study suggest that recipient households incur little to no cost with an easy
travel experience to access their cash. These results help explain the high success rate of completed
payments during the first 2 years of the program’s operations, with 98 percent of households in the
study receiving all their payments during this time. Almost every recipient walks to the pay point (97
percent), with under 1 percent reporting that they paid any money for travel. Most recipients do not
walk far to collect their payment; the median round trip travel time is under 10 minutes. Upon arrival,
recipients wait on average less than 17 minutes to receive their payment. Less than 10 percent of
recipients report ever having to make multiple trips to receive a single payment. Last, 93 percent report
that they generally feel safe after collecting money from the pay point. Therefore, it appears that pay
points are appropriately located, easily accessible, quick, and reliable.
Almost all beneficiary households (96 percent) report that recipients regularly pick up the payments
instead of using family members or friends. Over 90 percent of recipients have identified a
representative, usually a family member or relative, to pick up payments if they are unable to. Thirty
20
percent of the recipients report that they have used their representative at least once. This procedure is
consistent with the instructions in the program’s operations manual.
Notifications: Nearly all recipients are happy with the payment method and notification process; only 4
percent of recipients report being dissatisfied. The most common recommendation for a better method
of payment is door-to-door delivery, indicating that modifications to the program are not necessary for
continued satisfaction with payment delivery.
A majority of households are informed about payments by CWAC members (74 percent), with the rest
hearing about payments through family members (7 percent), pay point staff (5 percent), community
leaders (5 percent), and other community members (5 percent).
Solicitations: Recipients rarely report solicitations, and nearly all recipients are happy with program
staff. Although 8 percent of households report that community members request money from them,
less than 1 percent report any requests from pay point staff or actually paying any amount of money to
any party. The recipients express satisfaction with both the pay point staff and the CGP representatives
(97 percent satisfied).
On-Time Payments: Overall, payments during the 2-year period have been consistently on time for all
three districts. Payments are scheduled bimonthly, so we expect the average time between payments to
be about 60 days. This is supported with the district data, which report an average of 59 days between
payments over the course of 13 disbursements. During the 2 years of implementation, Shangombo was
the only district to report missing a payment, and, therefore, a double payment was made. Over 90
percent of respondents report receiving a payment in the 3 months prior to the survey.
Program Understanding
Recipients demonstrate a mixed understanding of the policies for the cash transfer program. This
knowledge is important because it affects their expectations and behavior. Recipients were asked
various questions regarding their understanding of the program with respect to eligibility requirements,
funding sources, and resources for complaints.
Eligibility: Seventy-five percent understand that they are eligible for the program because they have a
child under 5 years old. The rest believe that they are eligible because they care for orphans (18 percent)
or are very poor (7 percent). Almost everyone believes that the eligibility criteria are fair (97 percent),
although this is not surprising because all respondents are actual program beneficiaries.
Most recipients believe that they will receive the cash transfers for 5 or more years (84 percent).
However, there is some misunderstanding about what is required to continue to receive payments.
Although the cash transfer is unconditional, almost 90 percent of recipients report having to follow
requirements to keep receiving payments. Providing adequate food and nutrition to their children and
keeping their children clean represent the most commonly perceived conditions. A majority of recipients
(85 percent) believe that families can be kicked out of the program for not following the stipulations of
the program. Roughly 80 percent of households who think that there are conditions also report that
beneficiary households are monitored to see whether they are following the rules.
21
Funding Sources: Recipients have a good understanding of where program funds originate. Half of all
recipients attribute the funding to the Government of Zambia, an additional 18 percent report the
MCDMCH specifically, and 22 percent say from a foreign NGO or donor.
Complaints: The Ministry has procedures in place for recipients to lodge grievances. Recipients seem to
understand that there is a system, but it is not clear whether they understand the process. Almost 75
percent report that there is someone to whom they can report program issues, although roughly the
same percentage believe that concerns are to be reported to the CWAC members. Only 18 households
(less than 2 percent) have contacted someone, and the main reported problems concern missing or
expired payments.
22
VI. Consumption Expenditures
The conceptual framework suggests that the primary direct impact of the CGP will be on the
consumption spending behavior of recipient households, so we expect to see the most important
impacts of the program on levels of spending, with relatively higher impacts on items that are more
sensitive to income. Table 6.1 shows the impact estimates for total per capita expenditure (row 1) and
then impacts on per capita spending on other consumption items. The CGP has increased total per
capita consumption spending by ZMW 15.18 per month, which is more than the per capita value of the
transfer. Thus, as expected among very poor households, almost all the income from the program is
consumed.
The subsequent rows of Table 6.1 show the distribution of the increased spending by category. The
majority of the increased spending goes to food (ZMW 11.60), which is 76 percent of additional
spending, followed by health and hygiene (ZMW 1.08) at 7 percent, clothing at 6 percent, and
transportation/communication at 6 percent. In contrast, there is no program impact on education,
domestic items, or alcohol/tobacco.
Table 6.1: Impact of CGP on Consumption Expenditure
Program Impact Baseline 24-Month
Treatment
24-Month
Control
Total 15.18 46.56 67.04 48.59
(5.07)
Food 11.60 34.45 50.16 35.85
(4.76)
Clothing 0.93 1.47 2.42 1.50
(5.71)
Education 0.10 0.49 1.19 0.99
(0.34)
Health 1.08 2.60 4.13 2.89
(4.22)
Domestic 0.53 6.11 6.40 5.64
(0.81)
Transport/Communication 0.86 0.91 2.23 1.29
(2.32)
Other -0.01 0.13 0.23 0.18
(-0.11)
Alcohol, Tobacco 0.09 0.41 0.29 0.26
(0.68)
N 4594
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics
clustered at the CWAC level are in parentheses. Bold indicates that they are significant at p < .05. All
estimations control for household size, recipient age, education and marital status, districts, household
demographic composition, and a vector of cluster-level prices.
23
Table 6.2 breaks down the program impacts by detailed food groups. The overall increase in food
spending is ZMW 11.60 as reported in Table 6.1; the largest share goes to cereals (ZMW 4.54), followed
by meats, including poultry and fish (ZMW 2.44), followed by fats such as cooking oil (ZMW 1.76) and
then sugars (ZMW 1.28). There is a clear shift away from roots and tubers (primarily cassava) and
toward protein (dairy, meats), indicating a possible improvement in diet diversity among CGP recipients.
Table 6.2: Impact of CGP on Food Expenditure
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Cereals 4.54 11.61 15.54 9.95
(3.26)
Tubers -0.924 4.96 4.56 4.93
(-1.25)
Pulses 1.22 0.94 2.00 0.77
(4.98)
Meats 2.44 6.78 11.43 7.91
(3.08)
Fruits, Veg 0.49 7.03 8.86 8.89
(0.56)
Dairy 0.76 0.88 1.27 0.48
(3.55)
Baby Foods 0.02 0.01 0.03 0.01
(0.78)
Sugars 1.28 0.79 2.61 0.98
(7.80)
Fats, Oil, Other 1.76 1.45 3.87 1.93
(6.13)
N 4594
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-
statistics clustered at the CWAC level are in parentheses. Bold indicates that they are
significant at p < .05. All estimations control for household size, recipient age, education and
marital status, districts, household demographic composition and a vector of cluster-level
prices.
Impacts by Household Size
The CGP provides the same-size transfer to a household regardless of the household’s size. CGP
households vary in size, with roughly half the households having 5 or fewer members (1,218
households) and the other half having 6 or more members (1,238 households). Therefore, the value of
the transfer per capita within a household greatly varies and could lead to differential program impacts.
We investigate the possibility of differential program impacts by household size, comparing smaller
households (5 or fewer members) with larger households (6 or more members). Throughout this report
we provide impacts by household size only when a difference exists. We begin by comparing the
demographic profile of smaller and larger households in the study sample. Table 6.3 shows that smaller
24
households are richer (in terms of total per capita expenditure) and more frequently found in Kalabo (39
percent). The demographic composition of smaller households is also different, with a larger share of
members ages 0–5 (41 percent) and a larger share (35 percent) of adults ages 19 to 35 year old (prime-
age). In contrast, larger households have a smaller share of pre-school children (11 percent) and a much
larger share of primary school-age children (28 percent). Larger households have heads that are slightly
older and much more likely to be married (81 percent). Thus it appears that larger households in our
sample are slightly further along in the life-cycle relative to smaller households.
Table 6.3: Mean Household Characteristics by Size of Household
All Size <= 5 Size > 5
Number of residents 5.69 4.03 7.42
Total expenditure per capita (ZMW) 46.40 55.29 37.20
Kalabo 0.33 0.39 0.27
Shangombo 0.33 0.28 0.39
Kaputa 0.33 0.33 0.34
Demographic composition
Share 0–5 years 0.36 0.41 0.30
Share 6–12 years 0.19 0.11 0.28
Share 13–18 years 0.08 0.05 0.12
Share 19–35 years 0.26 0.35 0.18
Share 36–55 years 0.09 0.06 0.11
Share 56+ 0.02 0.02 0.02
Head’s characteristics
Age 29.85 26.81 32.99
Years of schooling 4.06 4.18 3.93
Married 0.72 0.64 0.81
Never married 0.11 0.16 0.05
Widow 0.06 0.06 0.07
Divorce 0.07 0.09 0.05
N 2519 1281 1238
Table 6.4 shows impacts on total expenditure and broad groups by large and small households. Not
surprisingly given the flat transfer, impacts on total expenditure are double the size for small households
as they are for larger households, and this pattern also holds for the impacts on food and clothing.
However, there are now significant impacts on education spending among large households (ZMW 0.61)
and no impacts among small households. This result is consistent with the demographic profile of larger
households, which contain proportionately more school age children relative to smaller households (see
Table 6.3). The impact of the CGP is much larger among smaller households for health spending (ZMW
1.58), which is consistent with the larger proportion of very young children in smaller households. The
impact on transportation and communication (ZMW 1.39) spending is also over 4 times the size in
smaller households as it is in larger households.
25
Table 6.4: Impact of CGP on Consumption Expenditure by Household Size
Size <= 5 Size > 5
Baseline Mean Program Impact Baseline Mean Program Impact
Total 55.40 20.37 37.42 10.10
(4.40) (3.38)
Food 41.36 15.17 27.30 8.16
(3.99) (3.56)
Clothing 1.77 1.22 1.15 0.61
(5.24) (4.10)
Education 0.25 -0.25 0.74 0.44
(-0.51) (2.10)
Health 3.25 1.58 1.93 0.57
(3.88) (2.19)
Domestic 7.36 1.11 4.81 -0.02
(1.15) (-0.03)
Transport/Comm 0.84 1.38 0.98 0.34
(2.44) (0.62)
Other 0.16 -0.04 0.10 0.03
(-0.24) (0.48)
Alcohol, Tobacco 0.41 0.20 0.41 -0.03
(1.29) (-0.17)
N 2306 2288
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics
clustered at the CWAC level are in parentheses. Bold indicates that they are significant at p < .05. All
estimations control for household size, recipient age, education and marital status, districts, household
demographic composition and a vector of cluster-level prices.
Table 6.5 presents program impacts on food spending by household size. Here also there are some
interesting differences in terms of the composition of food spending that the CGP has impacted in small
and large households. In smaller households, the impact of the CGP on food is concentrated on cereals
(where 45 percent of the impact on food is derived) followed by meat (15 percent), fats (14 percent),
and pulses (13 percent). However among larger households, the impact of the grant on food is driven by
meats (32 percent) and then cereals (30 percent). Again these distinct patterns are likely linked to the
differences in household demographic composition between the two types of households; smaller
households have a larger share of pre-school children who eat more cereal, while larger households
have a greater share of school-age children (including teenagers) who eat more meats.
26
Table 6.5: Impact of CGP on Food Expenditure by Household Size
Size <= 5 Size > 5
Baseline Mean Program Impact Baseline Mean Program Impact
Cereals 13.83 6.78 9.32 2.44
(3.46) (1.91)
Tubers 5.63 -0.80 4.26 -1.05
(-0.68) (-1.63)
Pulses 1.03 1.90 0.85 0.59
(4.42) (2.75)
Meats 0.85 2.21 4.99 2.58
(1.94) (3.33)
Fruits, Veg 8.59 0.54 5.43 0.42
(0.39) (0.60)
Dairy 1.14 0.74 0.60 0.76
(2.07) (4.38)
Baby Foods 0.01 0.03 0.01 0.01
(0.70) (0.71)
Sugars 0.89 1.58 0.68 1.01
(5.63) (6.22)
Fats, Oil, Other 1.73 2.17 1.16 1.40
(4.85) (4.97)
N 2306 2288
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics
clustered at the CWAC level are in parentheses. Bold indicates that they are significant at p < .05. All
estimations control for household size, recipient age, education and marital status, districts, household
demographic composition and a vector of cluster-level prices.
Actual Versus Predicted Impacts
Using baseline data, we had predicted the impact of the CGP transfer on the composition of overall and
food spending under the assumption that households would treat the money from the CGP the same as
they would any other source of income. Table 6.6 compares the share of the transfer allocated to the
different spending items as well as the share that we predicted from the baseline data assuming only an
income effect of the program. In general terms, the predictions are most accurate for food in that over
three-fourths of spending from the CGP income is devoted to food, as we predicted. However we note
some interesting and noteworthy differences between the two columns. In particular, more of the
transfer (than predicted at baseline) is devoted to clothing, health, and transportation/communication
and less (than predicted) to domestic items. This seems to be consistent with qualitative feedback from
the field as well as the results from the operational data, which suggest that recipients believe that they
must use the money to clothe and feed their children.
27
Table 6.6: Comparison of Actual and Predicted Impacts on Spending Groups
Actual Impact Predicted Impact at
Baseline
Food 0.764 0.781
Clothing 0.061 0.025
Education 0.007 0.007
Health 0.071 0.050
Domestic 0.035 0.074
Transport/Communication 0.057 0.044
Other -0.001 0.004
Alcohol, Tobacco 0.006 0.015
Total 1.000 1.000
NOTE: Numbers are the share of the total transfer allocated to each spending item. Column 1
is the actual share at follow-up; column 2 is the predicted share estimated from baseline data.
Table 6.7 provides a similar comparison for the composition of food spending. The biggest surprise here
is the large discrepancy between the ex ante prediction of the share devoted to roots and tubers (0.14)
and the actual share (which is a decline of 0.08 but not statistically different from 0). Instead, a larger
share (than predicted) is devoted to cereals, sugars, fats, dairy, and pulses. This finding could be due to
the greater level of attrition in Kaputa than in the other two districts because people in Kaputa mostly
eat cassava, a tuber, but maize, a cereal, in the other two districts. Clearly CGP households now enjoy
both higher levels of overall consumption and more diet diversity in terms of increased consumption of
dairy and meat. However, there is also a significant increase in sugars, oils, and fats, so that not all the
increase in food consumption may be healthy, although among this highly food insecure population,
these increases in fats and sugars probably enhance diet diversity and improve nutritional intake.
Table 6.7: Comparison of Actual and Predicted Impacts on Food Spending
Actual Impact Predicted Impact at Baseline
Cereals 0.391 0.339
Tubers -0.080 0.142
Pulses 0.105 0.030
Meats 0.211 0.236
Fruits, Veg 0.043 0.182
Dairy 0.065 0.027
Baby Foods 0.002 0.000
Sugars 0.111 0.039
Fats, Oil, Other 0.152 0.005
Total 1.000 1.000
NOTE: Numbers are the share of the total transfer allocated to each spending item. Column 1 is
the actual share at follow-up; column 2 is the predicted share estimated from baseline data.
28
VII. Poverty and Food Security
Earlier in this report we showed that the CGP has a significant impact in raising the average consumption
level of households. In this chapter, we provide estimates of the program’s impact on measures of
poverty and food security. Figure 7.1 compares the distribution of per capita monthly consumption
expenditure between the two arms in each period; the vertical line is the severe poverty line as defined
by the Central Statistics Office in 2010 (ZMW 96.37) inflated to 2012 units (in these figures we drop the
top 1 percentile for ease of exposition). Individuals to the left of the line are in extreme poverty. In 2010,
the two distributions are almost identical, and most important, the same proportion of households (96
percent) in treatment and control samples are below the severe poverty line. In 2012, however, the
distribution of per capita expenditure among treatment households has clearly shifted to the right
relative to control households, and it now appears as if fewer households (91 percent) in the treatment
arm are below the severe poverty line compared to 96 percent in the control group.
Table 7.1 provides more details on the impact of the CGP on the three commonly used FGT poverty
indicators, the headcount, poverty gap, and squared poverty gap, using both the severe and the
moderate poverty lines. In column 1, we provide the simple (unadjusted) impact estimates. These, as
well as the means in the table, are weighted by household size to be representative of the population of
individuals living in beneficiary households. Beginning with the severe poverty line, the program reduces
the headcount rate by 5.4 percentage points; however, the largest program impacts are found for the
0.0
05
.01
.015
.02
0 50 100 150 200Per Capita Monthly Expenditure
Treatment Control
20100
.005
.01
.015
.02
0 50 100 150 200 250Per Capita Monthly Expenditure
Treatment Control
2012
Figure 7.1: Distribution of Expenditures
29
poverty gap (10.9percentage points) and squared poverty gap (10.8 percentage points), which account
for the distribution of individuals below the line rather than whether individuals moved above the line.
For programs that target people at the very bottom of the income distribution, these last two indicators
are better measures of changes in welfare because it is highly unlikely for a program to provide
sufficient funds to lift people at the very bottom of the distribution to above the poverty line. However,
a significant positive movement below the line will show up in the poverty gap and squared poverty gap
indicators. Thus, this pattern of results is evidence of both the highly successful targeting approach of
the CGP as well as its impact on welfare.
Virtually all CGP recipients are below the moderate poverty line (99 percent), and the impact of the
program on the poverty headcount using the moderate poverty line, although statistically significant, is
very tiny at 1.7 percentage points. However, the impacts on the poverty gap and squared poverty gap
continue to be large simply because these indicators account for the distribution of individuals below
the line. Notice that among the control group, there is also a clear trend of improvement in terms of the
poverty gap and squared poverty gap, although the gains in monetary welfare among the CGP
participants is an order of magnitude larger in terms of percentage change from baseline.
Table 7.1: Impact of CGP on Poverty Indicators
Means
Percent Change From
Baseline
Program
Impact Baseline Treated Control Treated Control
Severe Poverty Line
Headcount -0.054 0.958 0.906 0.960 -5.43 0.21
(-3.71)
Poverty Gap -0.109 0.632 0.483 0.607 -23.58 -3.96
(-4.54)
Sq. Poverty Gap -0.108 0.456 0.293 0.420 -35.75 -7.89
(-4.19)
Moderate Poverty Line
Headcount -0.017 0.989 0.974 0.991 -1.52 0.20
(-2.76)
Poverty Gap -0.090 0.743 0.627 0.727 -15.61 -2.15
(-4.71)
Sq. Poverty Gap -0.102 0.593 0.449 0.567 -24.28 -4.38
(-4.49)
N 4815
NOTE: Program impacts are raw difference-in-differences with cluster robust t-statistics in parentheses. All estimates are
weighted by household size and corrected for attrition bias.
Food Security
One of the goals of the CGP is to improve the food security of beneficiary households and specifically
increase the percentage of households eating two or more meals per day. As stated earlier, the program
has large impacts on consumption, with over 75 percent of additional expenditures going toward food
consumption. We find that these additional expenditures on food translate to greater food security, a
finding consistent with our predictions conducted at baseline. Table 7.2 and Figure 7.2 show the impacts
30
of the program on several food security indicators. The CGP increases the percentage of households
eating two or more meals per day by 8 percentage points, with almost everyone eating two or more
meals per day (97 percent). Although the difference between the treatment and control groups is only 8
percentage points, a possible ceiling effect limits the measurement of the program’s impact on this
indicator because the indicator has almost topped out and reached its limit with only 3 percent
remaining in the treatment group who eat fewer than two meals per day.
Fortunately, other indicators, such as the Food and Nutrition Technical Assistance Project (FANTA) food
security score, provide greater depth to the program’s impact. FANTA is a measure of a household’s
food insecurity, with greater values indicating more food insecurity. We find that the program reduces a
household’s food insecurity score by 2.5 points, a 20 percent decrease from the control group’s score.
The program increases the number of households that are not severely food insecure by 18 percentage
points (36 percent in the treatment group versus 16 percent in the control group), a 113 percent
improvement over the control group. The CGP has a strong impact on perceptions of food security.
Twice as many CGP households (71 percent) as control households (35 percent) do not consider
themselves very poor. Five times more CGP households (60 percent) than control households (12
percent) report being better off now than they were 12 months ago. Thus, it appears that the CGP
improves household food security with strong impacts on one of the primary goals of the program, to
increase the number of households eating two or more meals per day.
Figure 7.2: Food Security Indicators by Treatment Status and Time
Eats more than
one meal a day
Not severely
food insecure
Does not
consider itself
very poor
Better off than
12 months ago
0
10
20
30
40
50
60
70
80
90
100
Pe
rce
nta
ge
24 month followup treatment
Baseline Treatment
24 month followup control
Baseline control
31
Table 7.2: Food Security Indicators, by CGP Treatment
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Eats more than one meal a day 0.079 0.78 0.97 0.89
(4.02)
Ate meat/fish => 5 times in last month 0.006 0.31 0.32 0.27
(0.11)
Ate vegetables => 5 times in last month -0.006 0.61 0.74 0.74
(-0.09)
Food security scale 2.498 15.10 9.63 12.36
(4.23)
Is not severely food insecure 0.177 0.10 0.36 0.16
(4.00)
Does not consider itself very poor 0.305 0.41 0.71 0.35
(-5.78)
Better off than 12 months ago 0.453 0.10 0.60 0.12
(10.51)
N 4,549 2,249 1,153 1,145
NOTE: Estimations use difference-in-difference modeling among panel households. Cluster robust t-statistics are n
parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient age,
education and marital status, districts, and a vector of cluster-level prices. All estimates are corrected for attrition bias.
Table 7.3 reports impacts for these same indicators by household size. Keeping in mind that these are
mostly self-assessments of welfare, we see a distinct pattern of larger households reporting larger
positive impacts of the CGP on their self-assessed welfare. For example, larger households are more
likely to report that they are better off than 12 months ago (47 versus 43 percent), that they do not
consider themselves very poor (33 versus 28 percent), and that they are not severely food insecure (21
versus 14 percent). This appears somewhat contradictory to the fact that the value of the grant (in per
capita terms) is much larger for smaller households and the actual impact of the CGP on monetary
welfare is larger in smaller households. In contrast, larger households are actually much poorer on a per
capita basis, so the grant has a larger impact on psychological welfare, which in turn shows up in these
self-reports of well-being. The relationship between psychological and material welfare is gaining
increasing attention in the literature on poverty.12
12
Grant, C. (2005). Insights on development from the economics of happiness. World Bank Research Observer,
20(2), 201–231.
32
Table 7.3: Impacts on Food Security Indicators by Household Size
Size <= 5 Size > 5
Baseline Mean Program
Impact
Baseline
Mean Program Impact
Eats more than one meal a day 0.770 0.085 0.805 0.069
(3.68) (3.02)
Ate meat/fish =>5 times last month 0.296 -0.013 0.334 0.021
(-0.21) (0.33)
Ate vegetables >= 5 times in last week 0.602 0.014 0.626 -0.027
(0.19) (-0.44)
Food security scale 15.11 2.305 15.24 2.602
(3.14) (4.60)
Is not severely food insecure 0.115 0.145 0.0792 0.211
(2.64) (4.78)
Does not considers itself very poor 0.393 0.284 0.441 0.326
(4.12) (6.02)
Better off than 12 months ago 0.0957 0.434 0.0981 0.471
(7.79) (9.44)
N 2306 2288
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices.
Diet Diversity
Repetitive diets are typically common in food insecure regions in many parts of the world. These add to
the load of undernourishment, mostly insufficient micronutrient consumption. An essential element to
food-based approaches involves dietary diversification or consumption of a wide variety of foods across
nutritionally distinct food groups. Increased dietary diversity is associated with increased household
food access as well as individual probability of adequate micronutrient intake. As an indicator of food
access, dietary diversity is defined as the number of individual food stuffs or food groups consumed over
a given reference period.13
A standardized tool for measuring dietary diversity has been developed by the Food and Agricultural
Organization of the United Nations (FAO).This tool can be administered at household and individual
levels. Using an open recall method, the tool gathers information on all the foods and beverages
consumed over the previous 24 hours by the household or individual. Food and beverages declared by
the respondent(s) are then recorded into one of 16 standardized food groups.
Most often, dietary diversity is measured by counting the number of food groups rather than food items
consumed. We adopt the same approach in this study, with 2- and 4-week reference periods for a
number of selected food groups. Households were asked to recall all the foods eaten and beverages
taken in the 2 and 4 weeks prior to the interview. We use this reference period because it offers a better
clue of the habitual diet of households.
The type of dietary diversity scores calculated are household dietary diversity scores (HDDS).These at
least portray a household’s capability to consume assorted food stuffs. By means of the data collected
13
Hoddinott, J., & Yohannes, Y. (2002). Dietary diversity as a food security indicator. Washington, DC: Academy for Educational
Development, Food and Nutrition Technical Assistance Project.
33
from the 2- and 4-week dietary reference period, we calculated the HDDS by using the FAO and FANTA
guidelines for measuring household and individual dietary diversity. The HDDS were calculated on the
basis of the 12 selected food groups consumed over the previous 2 and 4 weeks because the
consumption of certain food items belonging to the cereals food group were in the 2- and 4-week
reference periods, respectively. We awarded a point to each food group consumed over the reference
period and then calculated the sums of all points for the dietary diversity score for each household.
Following the FAO guidelines measuring household dietary diversity, the HDDS uses 12 food groups:
cereals, roots and tubers, vegetables, fruits, meat, eggs, fish, pulses and legumes, fats and oil, sugar and
sweets, milk and other milk product, and spices and beverages.
At baseline, vegetables, cereals, and fish were consumed on average more than any other food group
(Table 7.4). Among the least consumed food groups were eggs, milk and other milk products, and spices
and beverages. The 24-month follow-up indicates an increase in the consumption of the majority listed
food groups. On average, the highest increase in consumption is for fats and cooking oil (35 percent),
followed by sugars and sweets (31 percent), pulses and legumes (22 percent), and meat products (17
percent). Further analysis shows that within and across CGP districts, Kaputa has more households
consuming roots and tubers than cereals. This is contrary to what we expected because maize cereal is
believed to be the main staple in Zambia.
Table 7.4: Distribution of Food Groups Consumed
Food Group Baseline Survey
24-Month Follow-Up Survey
24-Month Treatment 24-Month Control Difference
(T-C)
(1) (2) (3) (4)
---------------------------- Percent (%) --------------------------
Cereals 81 98 93 5
Roots and tubers 49 46 46 0
Vegetables 91 97 96 1
Fruits 38 55 53 2
Meat 21 45 28 17
Eggs 4 9 3 6
Fish 77 85 79 6
Pulses and legumes 30 48 26 22
Fats and cooking oil 27 72 38 35
Milk and other milk products 24 30 19 10
Sugars and sweets 24 63 32 31
Spices and beverages 15 28 18 10
Number of Observations 2,517 1,148 1,142
On average, the overall HDDS in the baseline was about 4.78. In the 24-month follow-up, the HDDS is
6.73 and 5.30 in the treatment and control groups, respectively (Table 7.5). CGP households consumed
one more food group than their counterpart non-CGP households.
34
Table 7.5: Mean Household Dietary Diversity Scores at Baseline and Follow-Up
Variable Baseline
24-Month
Treatment
24-Month
Control Difference
HDDS 4.78 6.73 5.30 1.43
Observations 2,519 1,153 1,145
35
XIII. Young Child Outcomes
In this section, we report program impacts on a series of young child indictors covering health, use of
services, nutritional status, and early childhood development. We remind the reader that most of these
are second-round effects in that they are not affected directly by the cash transfer but require a series
of behavioral responses by the household induced by the income effect of the cash transfer in order to
change. For example, nutritional status is affected by caregiving behaviors, caloric intake, and sanitation.
For the CGP to affect nutritional status, it must induce a change in feeding practices or the disease
environment of the household. In the baseline report, we presented some predictions of where we
might expect to see impacts of the CGP. We reproduce this information here (Figure 8.1) to give the
reader an idea of where we are likely to find effects of the CGP, assuming that recipients spend cash
from the program in the same way as they spend other sources of income. The dark bars in this figure
indicate effects that are likely to be statistically significant; the largest expected impact of the CGP is on
infant and young child feeding (IYCF), followed by certain components of the early childhood indicators,
incidence of diarrhea, and weight for height. It is useful to keep these predictions in mind as we go
through the observed impact estimates.
Figure 8.1: Predicted Impacts of the CGP
Morbidity
Table 8.1 shows impact estimates on the three main illnesses occurring among preschool children. We
see a strong program impact on the prevalence of diarrhea in the previous 2 weeks—a decline of 4.9
percentage points—and a somewhat smaller effect of 3.6 percentage points on acute respiratory illness
(cough), although not statistically different from zero. The strong effect on the prevalence of diarrhea is
consistent with our ex-ante predictions shown in the figure.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
Play
thin
gs
Early
Educatio
n
Adeq
uate
Care
EC
D In
dex
IYC
F
Diarrh
ea
Fev
er
AR
I
HA
Z
WH
Z
WA
Z
Eff
ect
Siz
e in
SD
Un
its
Predicted Effect of 1 SD Increase in Expenditure Per Capita on Young
Child Indicators
36
Table 8.1: Impacts on Morbidity Among Children 0–60 Months
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Diarrhea -0.049 0.185 0.0684 0.0925
(-2.38)
Fever -0.019 0.233 0.113 0.125
(-0.53)
Acute respiratory
illness -0.036 0.203 0.0511 0.0832
(-1.42)
N 7232
NOTES: Reference period for illnesses is 2 weeks. Estimation uses difference-in-difference
modeling among panel households. Cluster robust t-statistics are in parentheses. Bold
indicates that they are significant at p < .05. All estimations control for household size,
recipient age, education and marital status, districts, and a vector of cluster-level prices. All
estimates are corrected for attrition bias.
Use of Health Services
Table 8.2 shows impacts on the use of services, including the household’s possession of a birth
registration document for their children under age 5. This document is not exactly a health indicator but
is strongly related to assisted delivery, which itself is a key health service. The only statistically significant
program effect is for the treatment of acute respiratory illness (ARI) and indicates a reduction in curative
care for ARI among children in the program, an opposite effect from what we expect. Note that the ex-
ante analysis suggests no impacts on curative or preventive care-seeking behavior.
Table 8.2: Impacts on Use of Services Among Children 0–60 Months
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Sought preventive care (N=7135) -0.045 0.776 0.788 0.791
(-1.14)
Has birth registration document (N=7646) -0.063 0.402 0.238 0.251
(-0.79)
Sought care for diarrhea* (N=972) 0.039 0.744 0.798 0.796
(0.54)
Sought care for fever* (N=1293) 0.012 0.726 0.848 0.823
(0.16)
Sought care for ARI* (N=1005) -0.142 0.341 0.157 0.267
(-2.00)
* Only estimated on sample that reported this illness in the prior 2 weeks.
NOTE: Estimations use difference-in-difference modeling among panel households. Cluster robust t-statistics are in
parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient age,
education and marital status, districts, and a vector of cluster-level prices. All estimates are corrected for attrition bias.
Nutritional Status and Feeding
Table 8.3 shows impact results for the three commonly used anthropometric indicators of weight-for-
height (a short term measure of underweight), weight-for-age (undernutrition), and height-for-age
(chronic or long-term nutritional status), all measured in standard deviations or z-scores using the new
37
World Health Organization reference tables. The CGP has induced an improvement in the weight of
young children, with effects on weight-for-height and weight-for-age of about 0.12 standard deviation,
although these are just outside the levels of statistical significance. We investigate whether program
impacts are different for different age groups among young children and see a large effect on weight-
for-height among children ages 3 to 5.
The table also shows a large highly statistically significant impact of the CGP on IYCF—an increase of 22
percentage points or an 88 percent increase over the baseline mean. This result is consistent with the
consumption expenditure effects reported earlier, as well as the ex-ante predictions suggesting that the
CGP would have a strong impact on this indicator. Because feeding is an important determinant of
weight, we checked whether there are noticeable impacts of the CGP on weight among children 6 to 24
months but do not find any statistically significant effects.
Table 8.3: Impacts on Nutritional Status and Feeding
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Weight for age z-score (N=6825) 0.128 -0.902 -0.900 -0.963
(1.89)
Weight for height z-score (N=6157) 0.118 -0.180 -0.0961 -0.154
(1.74)
Height for age z-score (N=6155) 0.066 -1.416 -1.445 -1.491
(0.70)
Young child feeding (N=1983) 0.217 0.317 0.596 0.434
(3.54)
NOTE: Nutritional indicators are reported for children 0 to 60 months; child feeding are reported for children 6
to 24 months as recommended in the ZDHS. Estimations use difference-in-difference modeling among panel
households. Cluster robust t-statistics are in parentheses. Bold indicates that they are significant at p < .05. All
estimations control for household size, recipient age, education and marital status, districts, and a vector of
cluster-level prices. All estimates are corrected for attrition bias.
Early Childhood Development
As we reported in the baseline report, an innovative aspect of the questionnaire we administered is the
inclusion of the newly released early childhood development (ECD) module developed and tested by
UNICEF as part of its global MICS 4 Program. We administered this module to children ages 3 through 7
in our sample and constructed six MICS recommended indicators from the MICS Child Development
Indicator list (indicators 6.1 and 6.3–6.7).14 Support to Learning measures whether an adult played with
the child, counted, named or drew things with the child, sang songs or told stories to the child, read
books or looked at pictures with the child, or took the child outside the compound. Learning Materials
refers to whether the child possesses at least three books or whether the child plays with homemade or
store-bought toys or objects around the home, such as pots, bowls, rocks, or sticks. Adequate Care
measures whether the child was ever left alone for more than 1 hour or left in the care of someone less
than 10 years old. School Attendance includes any sort of formal program, including preschool and
14
See http://www.childinfo.org/mics4_tools.html.
38
daycare. Finally, the ECD Index is a 10-item scale that covers four developmental domains: physical
(both gross and fine motor), language and cognition, socio-emotional, and approaches to learning.
Table 8.4 shows the marginal probability impact estimates on these six ECD indicators plus the overall
ECD Index Score. Households in the CGP show significantly higher support for learning and learning
materials as well as attendance at a formal educational program. The overall ECD Index Score has also
increases noticeably, although it remains just outside statistical significance. The ex-ante simulations
predicted a strong impact on playthings rather than support for learning or books/toys. It is interesting
to note the strong increase in the mean level of playthings and adequate care in both arms; in these two
cases, the absence of a control group would have suggested very strong program effects on these
indicators, highlighting the benefit of having an experimental control group with which to capture
overall trends over time in our indicators of interest.
Table 8.4: Impacts on Early Childhood Development Indicators
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Support to Learning (6.1) 0.126 0.432 0.307 0.225
(2.33)
Learning Materials: Books (6.3) 0.010 0.0148 0.027 0.011
(2.09)
Learning Materials: Playthings (6.4) -0.035 0.629 0.767 0.751
(-0.51)
Adequate Care (6.5) -0.003 0.307 0.620 0.647
(-0.04)
ECD Score 5+ (6.6) 0.070 0.558 0.610 0.572
(1.32)
ECD Index Score* 0.311 4.848 5.174 4.926
(1.62)
School Attendance (6.7) 0.041 0.224 0.154 0.136
(1.76)
N 5670
NOTES: All estimates are marginal probabilities from probit regression except those with *, which are OLS because
they are continuous instead of binary variables. Statistical significance at 10 percent or better is shown in bold. The
MICS indicator number is shown in parentheses beside the indicator name.
A key objective of the CGP is to ensure that infants and young children receive a healthy start to life. The
results presented here suggest that the program is meeting its goal. Specifically, children in beneficiary
households are less likely to be sick with diarrhea and have higher weight-for-height and weight-for-age
(although these two effects are just outside conventional levels of significance), and children ages 6 to
24 months are more likely to have the minimum recommended feeding. Children ages 3 to 7 years in
CGP households also have a better developmental environment, with greater support to learning and
more learning materials. These results are especially encouraging considering that this is a purely
demand-side intervention without any conditions attached to the receipt of the transfer and without
any explicit supply-side incentives to boost the use of services. For example, a recent meta-analysis of
39
the effectiveness of cash transfers on children’s nutritional status concluded that the impacts were close
to zero, underscoring the difficulty in moving an indicator such as height-for-age, which is determined by
a range of factors of which income is only one.15
15
Manley, J., Gitter, S., & Slavchevska, V. (2011). How effective are cash transfer programs at improving nutritional
status? (Working Paper No. 2010-8). Towson, MD: Towson University, Department of Economics.
40
IX. Children Over 5 Years Old
Although the CGP targets households with children under age 5, older children might benefit from living
in a household that receives the program, depending on how the money is spent. The conceptual
framework in section II demonstrates how the cash might have an impact on certain areas, such as
children’s material well-being, education, and health. At baseline, we ran simulations to predict where
we believed impacts were most likely to occur, based on the estimated elasticity of demand and
spending patterns. We concluded that material well-being would likely improve and that there could be
a small change in school attendance for older children, but we did not expect impacts for other older-
child-related indicators because the transfers were not expected to be spent in ways to affect these
outcomes. We investigate the effects of the CGP after 2 years on a number of outcomes in these areas
for children ages 5 to 17. As expected, we find large impacts on material well-being but none on
education or health. These results are supported by the spending patterns observed 24 months into the
program. Recipient households spend 6 percent of their additional money on clothing but less than 1
percent of their additional money on education, so the lack of results for education is not surprising.
Material Well-Being
The CGP has a large impact on children’s material well-being, indicating that recipients use some of the
transfer to purchase blankets, clothing, and shoes, items deemed necessary for supporting orphans and
vulnerable children.16 The material well-being indicator is a scale from 0 to 3; a child gets a point for
having a shared blanket, a second set of clothing, and shoes. At baseline, only 11 percent of the children
ages 6 to 17 had all three items. Two years later, 61 percent of the children in recipient households have
a blanket, a change of clothing, and shoes, whereas only 26 percent of the children in nonrecipient
households have all three items. The CGP increases children’s material well-being by 34 percentage
points. This impact is largely due to the increase in the number of children with shoes in recipient
households compared with those in nonrecipient households. Table 9.1 shows the impact of the
program on each item that makes up the material well-being scale. The program has an impact on both
shoes and blanket ownership, with shoes dominating this effect with a 33 percentage point increase (20
percentage point increase for blankets and 8 percentage point increase for clothing). A ceiling effect
occurs for clothing because 97 percent of children in recipient households and 89 percent of children in
nonrecipient households own a second set of clothing 2 years into the program. Therefore, there is little
room for recipient households to improve more than nonrecipients on this indicator, yet the difference
is still significant. This study asks about a second set of clothing, but perhaps children in recipient
households own more clothing than children in nonrecipient households, an indicator not captured
here.
16
The material well-being scale is a recommended indicator to measure care and support for orphaned and
vulnerable children. See UNICEF. (2005). Guide to monitoring and evaluation of the national response for children
orphaned and made vulnerable by HIV/AIDS. New York, NY: Author. Available at
http://www.measuredhs.com/hivdata/guides/ovcguide.pdf
41
Table 9.1: Child Needs Met at Ages 5–17, by CGP Treatment
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
All needs met 0.334 0.11 0.61 0.26
(5.47)
Child has shoes 0.331 0.14 0.62 0.29
(5.15)
Child has two sets of clothing 0.140 0.63 0.97 0.88
(4.47)
Child has a blanket 0.252 0.58 0.96 0.78
(6.04)
N 8.367 1,936 2,022
NOTE: Estimations use difference-in-difference modeling among panel households. Cluster robust t-statistics in
parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient age,
education and marital status, districts, and a vector of cluster-level prices.
Figure 9.1 shows the change from baseline to the 24-month follow-up for the treatment and control
groups on each material well-being indicator. Both groups improved during the 2-year period, but the
treatment group improved more than the control group as a result of CPG. We suspect that the control
group’s growth results from the bumper harvests that occurred during the study period and general
economic improvement of the country.
Figure 9.1: Material Needs Met by Treatment Status and Time
Shoes Clothing Blanket All 3
0
20
40
60
80
100
120
Pe
rce
nta
ge 24 month followup treatment
Baseline Treatment
24 month followup control
Baseline control
42
Education
We investigate education outcomes related to enrollment, on-time enrollment, and attendance for
children ages 6 to 16 and by gender. As predicted from a simulation with baseline data, we do not find
any overall impacts on education outcomes to the entire group. However, we find strong impacts on
these outcomes for children whose mothers are less educated. In other words, on average, the lower a
mother’s level of education completed, the greater the impact that CGP has on her children’s education.
Children living in a beneficiary household are 1 percentage point more likely to ever enroll in school and
2 percentage points more likely to enroll on time, for every year less of education their mother has.
These are statistically significant impacts of the program on the treatment group compared with the
control group. This result may sound counterintuitive because typically a mother’s education is
positively correlated with her children’s education. This relationship holds true in Zambia, too, including
with our sample, where at baseline, children of more educated mothers were more likely to be enrolled
in school and attending school regularly. So how do we explain these findings? Well-educated mothers
were already enrolling their children in school at baseline; therefore, the cash transfer has little
opportunity to improve how they act. However, it seems that the CGP enables or motivates less
educated mothers who did not enroll their child in school at baseline to change their actions and start
enrolling their child in school, leading to a program impact on education for children with less educated
mothers, but no impact on education for children with educated mothers because these children are
already enrolled and attending school.
Health
We investigate health outcomes for older children with respect to morbidity, treatment seeking, and
chronic illness. As predicted at baseline, we do not find any impacts on these health outcomes for
children over age 5. Illness is a rare event, with only 10 percent of the children reporting that they were
ill or injured in the previous 2 weeks. Of the 10 percent who reported being ill, 80 percent sought
treatment. This rate of treatment is up from baseline by roughly 20 percentage points, but it occurs
evenly in both the treatment and the control groups. Chronic illness is also an extremely rare event for
this group, with less than 1 percent of older children reporting a chronic illness at baseline. Although the
lack of impacts on health among older children was predicted at baseline, the expenditure analysis
shows that 7 percent of the transfer is spent on health (mostly user fees and drugs), although there is no
analogous increase in the use of curative care among older or younger children.
43
X. Women
Although the CGP is targeted toward children under age 5, because cash is in most cases given directly
to women, there is potential for impacts on women-level outcomes. As demonstrated in the conceptual
framework, these impacts depend on many factors, including power relations in households and
characteristics of women, such as how future looking they are in determining consumption patterns.
The following section explores trends and the impact of CGP on bargaining power, savings, future
outlook, and women’s health. Although we find significant impacts on women’s savings and future
outlook, we find no measurable impacts on decision making and health outcomes, with the exception of
self-rated measures. Lack of impact on decision making can be partially explained by upward trends in
indicators over the project cycle.
Bargaining Power
To explore bargaining power among sample households, we asked decision-making questions in nine
domains: children’s health, children’s schooling, spending of own income, spending of partner’s income,
major household purchases, daily household purchases, spending on children’s clothes and shoes, visits
to family and relatives, and own health. These questions were asked of one woman per household
(typically a mother or caregiver of a target child), and they allowed the respondent to answer whether a
decision is typically made by herself, by her partner, jointly, or by someone else in the household. To
explore impacts, we construct two indicators. The first is a count or summation, giving 1 point to each
time the woman indicates having sole decision-making power in a domain (ranges from 0 to 9). The
second is an index constructed by factor analysis, which weights indicators differently on the basis of
their variation within the sample and correlation between each other.
Table 10.1 shows the impact of the program on the count indicator and the index of sole decision
making. Results indicate that the program has no measurable impact on sole decision making, a finding
that remains unchanged even when we consider sole or joint decision making.
Table 10.1: Women’s Decision Making, by CGP Treatment
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Count indicator of sole decision making (9 domains) 0.205 3.93 4.46 4.37
(0.90)
Index of sole decision making (9 domains) 0.055 -0.07 0.08 0.06
(0.83)
N 4,498 2,257 1,115 1,126
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, and a vector of cluster-level prices.
However, as seen in Figure 10.1, there are notable increases over time for both treatment and control
groups in all decision-making domains. For example, the percentage of women responding that they
alone have decision-making power about their child’s health increases from approximately 56 percent to
44
approximately 70 percent in both treatment and control groups. Similar gains can be seen across other
decision-making domains, and it is possible that these overall trends may mask program impacts.
However, the lack of measurable impact indicates that transfers are seen as common household
resources and are not necessarily changing women’s bargaining power within households after 24
months.
Savings and Future Outlook
In Table 10.2, we investigate indicators of savings and future outlook as reported by the female
respondents answering bargaining-power questions for each household. Results indicate that at
baseline, approximately 16 percent of households had any saving in the previous 3 months. However, by
the 24-month follow-up, this percentage increased to 47 percent, while control households increased by
a smaller fraction to 22 percent. As expected, we find a large and significant program impact on any
savings and similarly on the amount of savings reported in ZMW. These impacts demonstrate that
households not only are using the transfer for immediate consumption but also are saving a portion of
the transfer. We also find significant impacts on future outlook. At baseline, 61 percent of households
believed that life would improve over the next 3 years, and this increases to 91 percent among
treatment households, and less so to 82 percent among control households.
0
10
20
30
40
50
60
70
80
Pe
rce
nta
ge
Figure 10.1: Women's Sole Decision Making, by Treatment Status and Time
24 month followup
treatment
Baseline Treatment
24 month followup
control
Baseline control
45
Table 10.2: Savings and Future Outlook, by CGP Treatment
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Any savings last 3 months 0.201 0.16 0.47 0.22
(3.42)
Log amount saved last 3 months (ZMW) 2.667 1.74 5.29 2.31
(5.45)
Believes life will be better next 3 years 0.115 0.61 0.91 0.82
(3.24)
N 4,498 2,253 1,112 1,125
NOTE: Estimations use difference-in-difference probit modeling and OLS modeling (log savings) among panel households.
Robust t-statistics clustered at the cluster level in parentheses. Bold indicates that they are significant at p < .05. All estimations
control for household size, recipient age, education and marital status, districts, and a vector of cluster-level prices.
Women’s Health
We investigate health outcomes for women age 18 and older with respect to morbidity in the previous 2
weeks, care seeking for illness, chronic illness in the previous 6 months, and self-reported health status.
We do not find any impacts on morbidity, care seeking, or chronic illness; however, these results are not
surprising given the emphasis of the program on children. In addition, the percentages reporting
morbidity and chronic illness at baseline were low (16 percent and 3 percent, respectively). We do find
impacts on self-rated health status. More specifically, women in treatment households are significantly
more likely to report “good health or better” and “very good health or better” than those in control
households. Although self-reported measures of health are subject to bias, this may be an indicator that
women are more optimistic about their health and economic situation in program households.
46
XI. Birth Outcomes
Although not a focus of the program, it is possible that the CGP impacts birth outcomes, including
antenatal care and skilled attendance at birth. Impact pathways include direct health care spending or
reallocation of resources through increases in women’s bargaining power. We examine a range of birth
outcomes, which are constructed as household-level averages for children born in the 24 months prior
to each survey (during the program period, and 24 months prior to baseline). This results in
approximately 1,634 households with baseline observations and 818 with 24-month follow-up
observations, for a total sample size of 2,514.
Table 11.1 shows results of our analysis and indicates that the program has no significant impacts across
all antenatal and skilled attendance indicators. For several indicators, including any antenatal care and
quality of care (tetanus vaccination, malaria prevention, and voluntary counseling and testing [VCT] for
HIV), it is unlikely we would observe impacts due to high baseline averages (e.g., 98 percent of the
sample report receiving any antenatal care; 92 percent report receiving a tetanus vaccination).
However, only 73 percent of the baseline sample report any antenatal care visit with a doctor or a nurse,
and only 24 percent report the first visit within the first trimester of pregnancy. Likewise, only 35
percent reported at baseline that the birth was attended by a doctor or a nurse. Unlike many of the
indicators examined in the report, there is no overall improvement in indicators from baseline to the 24-
month follow-up, indicating a lack of progress on these indicators. These are also roughly comparable to
the statistics found in the 2007 ZDHS, which collected information on births over the 5 years prior to the
survey. According to the ZDHS, antenatal care is nearly universal (97 percent); approximately 21 percent
of the sample visits within the first trimester, 59 percent of the sample completes at least the
recommended four visits, and 42 percent of births are attended by a doctor or a nurse. Overall lack of
impact means that increases on health expenditures are likely being allocated to young children and not
to pregnant mothers and that transfers are not inducing large shifts in bargaining power and
reallocation of resources. Future analysis will investigate possible heterogeneous effects by education of
recipient females or by health service provision within the community.
47
Table 11.1: Antenatal Care and Skilled Attendance, by CGP Status
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Received any antenatal care -0.003 0.98 0.99 0.99
(0.21)
Received antenatal care from doctor or nurse 0.024 0.73 0.73 0.74
(0.42)
Antenatal care within first trimester 0.009 0.24 0.27 0.20
(0.22)
At least four antenatal care visits -0.057 0.65 0.60 0.68
(0.88)
Tetanus vaccination during pregnancy -0.017 0.92 0.95 0.96
(0.60)
Malaria preventative medication during pregnancy -0.004 0.93 0.98 0.98
(0.19)
VCT during pregnancy -0.026 0.85 0.94 0.93
(0.69)
Birth attended by doctor or nurse 0.067 0.35 0.35 0.39
(1.34)
N 2,514 1,634 404 414
NOTE: Standard errors in parenthesis. Estimations use difference-in-difference probit modeling among panel
households. Robust t-statistics clustered at the CWAC level are in parentheses. Bold indicates that they are significant
at p < .05. All estimations control for household size, recipient age, education and marital status, districts, distance to
nearest health facility and a vector of cluster-level prices.
48
XII. Economic Impacts
CGP beneficiaries are poor, with limited options in terms of livelihoods and with few assets with which
to generate income. Beneficiary households have on average approximately half a hectare of
agricultural land, a couple of chickens, basic agricultural tools, and low levels of education, and they are
highly dependent on unskilled labor. A large majority of CGP beneficiaries are agricultural producers.
Almost 80 percent produce crops, and about half have some form of livestock. At the time of the
baseline survey, the beginning of the hunger season, home production accounted for almost 40 percent
of all food consumption. Most beneficiaries grow local maize, cassava, or rice, using traditional
technology and very low levels of modern inputs, and have little access to credit. About half of all
children work regularly on the family farm—including more than a third of those ages 5–10. Almost 40
percent of households have a non-agricultural enterprise at follow-up. Approximately 50 percent of
households, and a quarter of all adults, had some form of wage labor at baseline (mostly agricultural and
of a temporary nature), while 1/3 of households received private or public transfers.
Given the theory of change presented earlier, we expect the CGP to influence the livelihood activities of
beneficiary households. Two characteristics of the CGP program, compared with other programs in
Zambia and in the region, suggest a particularly large impact: the demographic profile, with relatively
more available household labor able to work, and the relatively large transfer size. We hypothesize that
the CGP will lead to an increase in household investment in agricultural input and labor use and
production and in the operation of household nonfarm business enterprises. Finally, the program could
affect the labor activities of individual household members, including the participation and intensity of
wage labor (agricultural and nonagricultural) and their own farm labor. We expect an increase in labor
dedicated to the beneficiaries’ own farms and a decrease in less desirable agricultural wage labor. It is
unclear the direction of impact on nonagricultural wage labor because this depends on labor market
conditions in the local economy, the relative returns between on- and off-farm labor, and household
domestic priorities.
As mentioned earlier, the effective per capita transfer is greater for small households than for large
households. Although we might expect the impact to be greater for smaller households, it is important
to keep in mind that small and large households are quite different in terms of demographic and
livelihood profiles. Large households are much poorer in terms of per capita levels of consumption, but
they have greater available productive resources in absolute terms. Although large households have a
bigger dependency ratio, they also have more available household labor to work in family agricultural
and nonagricultural businesses, as well as more male household members. Large and small households
employ the same productive activities with the same technology, but larger households operate at a
bigger scale. For example, at baseline they operated more land, used more productive inputs, produced
more output of maize and cassava, had over double the number of livestock holdings, and had a greater
number of livestock transactions.
Crop Production
We look at various dimensions of the productive process to ascertain whether households have
increased spending in agricultural activities, including crop production and crop input use. Overall, in
terms of these direct impacts on crop activity, we find positive and significant impacts on area of land
operated, overall crop expenditures, and specific expenditure on seeds, fertilizer, hired labor, and other
expenditures (Table 12.1). The CGP increases the amount of operated land by 0.18 hectares (a 34
percent increase from baseline), and the program has led to an increase of 18 percentage points in the
49
share of households with any input expenditure, from a baseline share of 23 percent (see Annex 4, Table
A4.1). This increase is larger among smaller households and includes spending on seeds, fertilizer, and
hired labor. Small beneficiary households spend ZMW 42 more on crop inputs than the corresponding
control households, including ZMW 15 on hired labor. This amounts to three times the value of the
baseline mean for overall spending, and four times for hired labor.
Table 12.1: Impact of CGP on Crop Input Use and Land Use (ZK)
Program
Impact
Baseline Program
Impact
Baseline Program
Impact
Baseline
All HH Size < 6 HH Size > 5
Operated land
(has)
0.179
(2.67)
0.49 0.162
(2.54)
0.43 0.197
(1.98)
0.56
Total crop exp 31.17
(2.97)
21.78 42.86
(5.14)
13.66 18.39
(1.12)
30.17
Exp seed 9.86
(4.41)
6.40 11.09
(4.94)
4.75 8.61
(2.65)
8.10
Exp hired labor 8.42
(1.45)
7.61 14.68
(4.19)
2.94 1.16
(0.11)
12.44
Exp pesticides 0.07
(0.40)
0.03 0.19
(1.13)
0.05 0.03
(0.13)
0
Exp fertilizer 7.60
(2.06)
1.40 8.92
(2.30)
0.66 6.50
(1.58)
2.16
Other crop exp 5.23
(2.00)
6.34 7.97
(2.59)
5.24 2.09
(0.59)
7.47
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices. Baseline
refers to baseline mean value of indicator shown in the preceding column.
We also see a positive impact on ownership of agricultural tools, but with two distinct patterns. For
implements widely available at baseline, such as axes and hoes (up to approximately 90 percent of
households at baseline), we see significant program impacts on the number of assets held (Table 12.2).
But for agricultural implements with low initial values (less than 10 percent at baseline), such as
hammers, shovels, and ploughs, we see a positive impact of between 3 to 4 percentage points on the
share of households now owning this equipment (Annex Table A4.2). In addition, the impact on
hammers, shovels, and ploughs is concentrated among larger households.
50
Table 12.2: Impact of CGP on agricultural implements (number)
Program
Impact
Baseline Program
Impact
Baseline Program
Impact
Baseline
All HH Size < 6 HH Size > 5
Axe 0.184
(2.43)
1.12 0.198
(2.41)
1.00 0.173
(1.74)
1.24
Pick 0.027
(1.15)
0.04 -0.006
(-0.22)
0.03 0.059
(2.12)
0.05
Hoe 0.296
(3.76)
1.54 0.214
(2.24)
1.34 0.388
(3.56)
1.75
Hammer 0.042
(2.16)
0.06 0.024
(1.12)
0.04 0.060
(2.06)
0.07
Shovel 0.027
(0.98)
0.06 -0.019
(-0.58)
0.04 0.075
(1.84)
0.09
Plough 0.033
(1.66)
0.07 0.021
(0.89)
0.06 0.052
(1.85)
0.09
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices. Baseline
refers to baseline mean value of indicator shown in the preceding column.
Does the increase in input use and tools lead to an increase in crop production? We focus primarily on
the three most important crops (maize, cassava, and rice) and aggregate all production by the value of
the total harvest.17 First, the program has facilitated some shifts in production in beneficiary households
compared with control households (Table 12.3). The share of (large) beneficiary households planting
maize has increased by 8 percentage points (from a baseline of 53 percent), whereas the share of small
beneficiary households planting rice has increased by 4 percentage points (from a baseline of 16
percent). The share of all households producing groundnuts, a relatively minor crop (5 percent at
baseline), has increased by 3 percentage points.
17
The value of total harvest is the product of harvest quantity and the median unit price; the latter is computed
from crop sales at the district level and, if missing, at the level of all three districts.
51
Table 12.3: Impact of CGP on Crop Production (share)
Program
Impact
Baseline Program
Impact
Baseline Program
Impact
Baseline
All HH Size < 6 HH Size > 5
Maize 0.049
(1.48)
0.55 0.020
(0.55)
0.53 0.081
(1.99)
0.58
Cassava -0.026
(-1.02)
0.26 -0.010
(-0.42)
0.21 -0.045
(-1.45)
0.31
Rice 0.031
(1.70)
0.16 0.039
(2.00)
0.17 0.019
(0.73)
0.15
Millet 0.010
(0.63)
0.06 0.010
(0.50)
0.07 -0.003
(-0.18)
0.06
groundnut 0.035
(3.35)
0.05 0.030
(2.83)
0.02 0.032
(2.11)
0.07
Sweet
potatoes
-0.000
(-0.03)
0.04 -0.007
(-0.92)
0.03 0.008
(0.89)
0.05
Sorghum 0.009
(0.91)
0.04 0.018
(1.22)
0.04 0.002
(0.16)
0.03
Other beans 0.009
(1.50)
0.01 0.012
(1.54)
0.01 0.007
(0.74)
0.02
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices. Baseline
refers to baseline mean value of indicator shown in the preceding column.
Aggregating all output by value, we find that the CGP has had a positive impact (at the 10 percent level)
in the value of all crops harvested—ZMW 146, approximately a 50 percent increase from baseline (Table
12.4). The impact rises to ZMW 182 for smaller households and is not significant for larger households.
We find few significant impacts, however, on the output of specific crops; the impact results on maize
are large and in the right direction but are not quite significant. The results are similar for rice, although
for small households, the positive impact is significant at 10 percent. Larger households had significantly
lower production of cassava (129 kg, from a baseline of 179 kg). This result is consistent with the decline
in consumption of tubers reported earlier.
Why is there a significant impact on the value of aggregate production, but little clear impact on specific
crops? It could be the result of a diffuse increase in production across crops. Differential crop price
increases between treatment and control households may have played a role, but we do not find any
systematic indication of this (see Annex 1). Note also that no production data have been collected on
fruits and vegetables, although the consumption model shows evidence of an increase in the share of
households consuming fruits and vegetables from home production. Finally, while households use more
inputs in production, they may not be using them in the most efficient manner—efficiency analysis is a
topic for further research.
52
Table 12.4: Impact of CGP on crop production (kg and 2012 ZMW)
Program
Impact
Baseline Program
Impact
Baseline Program
Impact
Baseline
All HH Size < 6 HH Size > 5
Maize 49.502
(1.62)
148.16 35.112
(1.54)
117.84 63.766
(1.25)
179.46
Cassava -68.142
(-1.67)
146.64 -16.958
(-0.51)
102.96 -129.226
(-2.05)
191.74
Rice 20.381
(1.32)
78.90 39.409
(1.79)
78.10 2.709
(0.16)
79.72
Millet 2.540
(0.90)
7.08 1.825
(0.55)
7.55 0.081
(0.03)
6.60
Groundnut 2.977
(0.63)
11.32 3.744
(1.37)
5.40 3.182
(0.38)
17.43
Sweet potato -6.406
(-1.05)
6.09 -3.683
(-0.61)
4.65 -8.077
(-0.88)
7.58
Sorghum 1.567
(0.53)
5.68 4.260
(0.88)
6.72 -1.233
(-0.61)
4.60
Other beans -0.531
(-0.84)
1.06 0.244
(0.34)
0.88 -0.977
(-0.82)
1.23
Value of
harvest
145.88
(1.95)
393.88 182.27
(2.40)
323.54 104.18
(1.04)
466.58
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices. Baseline
refers to baseline mean value of indicator shown in the preceding column.
Along with an increase in the value of crop production, a larger share of beneficiary households market
their crop production (an increase of 12 percentage points, from a baseline of 22 percent. The average
value of sales among all crop producing households is thus larger for beneficiary households (ZMW 82,
over double the baseline value of ZMW 76), although for larger households, the impact is significant only
at 10 percent (Table 12.5). The increase in market participation is driven by maize production in Kaputa
and by both maize and rice production in Kalabo. At the same time, the share of households consuming
some part of their harvest has increased by 6 percentage points (significant at the 10 percent level, as
seen in the last row of Table 12.5), which comes from increased groundnut and rice consumption of
home production (not shown).
53
Table 12.5: Impact of CGP on Agricultural Production
Progra
m
Impact
Baseline Program
Impact
Baseline Program
Impact
Baseline
All HH Size < 6 HH Size > 5
Value of sales
(ZMW)
81.52
(3.16)
75.77 86.27
(3.75)
63.59 73.80
(1.72)
88.35
% selling crops 0.120
(3.51)
0.22 0.144
(2.92)
0.20 0.092
(2.37)
0.24
Value of crops
consumed at home
(ZMW)
41.25
(1.49)
204.20 28.36
(1.03)
173.79 49.90
(1.36)
235.64
% of crops
consumed at home
0.059
(1.78)
0.76 0.063
(1.60)
0.73 0.057
(1.57)
0.80
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices. Baseline
refers to baseline mean value of indicator shown in the preceding column.
Livestock Production
The CGP has a positive impact on the ownership of a wide variety of livestock, both in the share of
households with livestock (a 21 percentage point increase overall, from 49 percent at baseline—Table
12.6) and in the total number of goats and poultry (an increase in 0.14 goats and 1.23 chickens, from
baseline values of 0.05 and 1.99, respectively—Table 12.7). Both small and large beneficiary households
have increased livestock ownership, but the impacts are particularly strong for large households. The
share of large households with livestock has increased 27 percentage points from a base of 55 percent
(compared with 16 percentage points for small households), including 5 and 21 percentage point
increases in the ownership of milk cows and chickens, respectively (compared with nonsignificant results
for small households). In terms of the number of livestock, the impact is more balanced between small
and larger households. Small household beneficiaries have obtained more goats and larger households,
more ducks. Overall, small households have accumulated more animals as measured in Tropical
Livestock Units (TLU),18 although significant only at the 10 percent level.
18
The TLU conversion factors are based on the average weight of animal species and aggregation of livestock into a
single index.
54
Table 12.6: Impact of CGP on Livestock Ownership (share)
Program
Impact
Baseline Program
Impact
Baseline Program
Impact
Baseline
All HH Size < 6 HH Size > 5
Milk cows 0.033
(1.74)
0.06 0.014
(0.75)
0.05 0.051
(2.15)
0.06
Other cattle 0.084
(4.02)
0.10 0.082
(3.30)
0.08 0.082
(3.02)
0.12
Chickens 0.154
(3.45)
0.41 0.097
(1.97)
0.36 0.214
(4.12)
0.47
Goats 0.036
(3.35)
0.02 0.034
(3.57)
0.01 0.035
(2.01)
0.03
Ducks 0.030
(2.78)
0.03 0.026
(2.08)
0.02 0.036
(2.06)
0.04
Total 0.209
(4.68)
0.49 0.155
(3.11)
0.43 0.266
(5.11)
0.55
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices. Baseline
refers to baseline mean value of indicator shown in the preceding column.
Table 12.7: Impact of CGP on Livestock Ownership (number)
Program
Impact
Baseline Program
Impact
Baseline Program
Impact
Baseline
All HH Size < 6 HH Size > 5
Milk cows -0.061
(-0.70)
0.21 0.019
(0.46)
0.09 -0.128
(-0.78)
0.33
Other cattle 0.263
(1.32)
0.45 0.227
(1.25)
0.33 0.269
(0.79)
0.57
Chickens 1.234
(3.28)
1.99 1.137
(2.77)
1.48 1.293
(2.57)
2.53
Goats 0.142
(4.31)
0.05 0.173
(3.52)
0.03 0.100
(2.45)
0.07
Ducks 0.198
(2.72)
0.12 0.150
(1.99)
0.10 0.258
(2.51)
0.15
Total (TLU) 0.138
(1.27)
0.37 0.165
(1.67)
0.24 0.102
(0.55)
0.50
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices. Baseline
refers to baseline mean value of indicator shown in the preceding column.
Further, overall, beneficiary households have a significantly larger volume of purchases and sales of
livestock compared with control households (Table 12.8). This increase in the volume is not significant
for smaller households; for larger households, the joint volume of sales (ZMW 73) and purchases (ZMW
55
110) is over twice as large as at baseline. In contrast to crop input use, no impact is found on
expenditures on inputs for livestock production, including vaccinations and other expenditures.
Table 12.8: Impact of CGP on Livestock Production (2012 ZMW)
Program
Impact
Baseline Program
Impact
Baseline Program
Impact
Baseline
All HH Size < 6 HH Size > 5
Total livestock
exp
-0.57
(-0.34)
1.16 -1.83
(-0.84)
0.48 1.11
(0.51)
1.88
Fodder exp 1.07
(1.61)
0.30 0.51
(1.82)
0.00 1,841
(1.25)
0.61
Vaccinations
exp
-0.52
(-0.81)
0.40 -1.04
(-1.08)
0.31 36
(0.09)
0.48
Other livestock
exp
-1.12
(-1.19)
0.47 -1.30
(-1.06)
0.16 -0.76
(-0.57)
0.79
Livestock
purchases
47.70
(2.93)
25.30 25.29
(1.20)
17.90 73.00
(3.02)
32.95
Livestock sales 55.56
(3.67)
34.66 13.43
(1.13)
13.94 109.51
(4.20)
56.07
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices. Baseline
refers to baseline mean value of indicator shown in the preceding column.
Nonfarm Business Activities
Beneficiary households of the CGP are significantly more likely to have a nonfarm enterprise (Table
12.9). The share of beneficiary households operating a nonfarm enterprise has increased by 17
percentage points compared with control households. Moreover, the program doubles the average
number of months in operation (reaching 2.8 months at follow-up), the value of total monthly revenue
(ZMW 184) and profit (ZMW 69), and the share of households owning business assets (5 percentage
points, reaching 12 percent at follow-up). The impacts are significant for both small and large
households, although the impact on ownership (7 percentage points) is significant only for large
households.
56
Table 12.9: Impact of CGP on Nonfarm Enterprise (NFE)
Program
Impact
Follow-up Program
Impact
Follow-up Program
Impact
Follow-up
All HH Size < 6 HH Size > 5
HH operates NFE 0.166
(4.42)
0.39 0.157
(3.60)
0.39 0.177
(4.50)
0.38
Months in operation 1.445
(4.44)
2.83 1.201
(3.38)
2.80 1.629
(4.23)
2.85
Total monthly
revenue (ZMW)
184.28
(4.43)
184.33 135.24
(3.77)
150.03 233.52
(3.65)
219.67
Total monthly profit
(ZMW)
69.08
(4.05)
81.87 55.13
(3.32)
72.98 81.24
(3.78)
91.03
Owned business
assets
0.0452
(2.51)
0.12 0.0238
(1.04)
0.13 0.0669
(3.22)
0.12
Value of owned
assets (ZMW)
196.64
(1.24)
134.63 17.18
(0.66)
46.63 342.05
(1.27)
225.06
N 2247 2247 1141 1141 1106 1106
NOTE: Estimations use single difference modeling. Robust t-statistics clustered at the CWAC level are in parentheses. Bold
indicates that they are significant at p < .05. All estimations control for household size, recipient age, education and marital
status, districts, and household demographic composition. Follow-up refers to follow-up mean value of indicator shown in the
preceding column.
Labor Supply
The changes in household economic activities brought on by the CGP necessarily imply changes in labor
activities of individual household members, the main input to household livelihoods, including wage
labor and agricultural and nonagricultural enterprises. Overall, we find a significant shift from
agricultural wage labor to family agricultural and nonagricultural businesses, which corresponds with the
increases in household-level economic activities brought on by receipt of the CGP transfer.
The CGP has led to a 9 percentage point decrease in the share of households with an adult engaged in
wage labor, from 59 percent at baseline (Table 12.10). The impact is much stronger for households with
females of working age—a decrease of 14 percentage points compared with no significant impact on
households with males of working age.19
19
In this analysis we join together permanent and temporary labor because only 3 percent of households have
access to permanent employment. Permanent workers typically refer to employees with paid leave entitlements in
jobs or work contracts of unlimited duration, including regular workers whose contracts last for 12 months and
over. Temporary employees usually have an expected duration of the main job of less than 1 year, carrying out
seasonal or casual labor.
57
Table 12.10: Participation in Any Labor Activity, HH Level
Program
Impact
Follow-up Program
Impact
Follow-up Program
Impact
Follow-up
All Males Females
Participation in
any labor
activity
-0.0913
(-2.79)
0.50 -0.0488
(-1.40)
0.44 -0.136
(-4.10)
0.40
N 2296 2296 1764 1764 2282 2282
NOTE: Estimations use single difference modeling. Robust t-statistics clustered at the CWAC level are in parentheses. Bold
indicates that they are significant at p < .05. All estimations control for household size, recipient age, education and marital
status, districts, and household demographic composition. Follow-up refers to follow-up mean value of indicator shown in the
preceding column.
In terms of types of employment, the reduction in wage labor has taken place primarily in agricultural
wage labor, with an 8 percentage point reduction for households with male labor and a 17 percentage
point reduction for households with female labor (Table 12.11). This result is expected because
agricultural wage labor is generally considered the least desirable labor—an activity of last resort. But
when liquidity is constrained, households may be obliged to overly depend on it. The CGP has also led to
a reduction in labor intensity in terms of days of agricultural wage labor, both overall (14 days fewer per
year) and for females (12 days fewer per year). The reduction in agricultural wage labor is also reflected
in the yearly value of household earnings, which is reduced by ZMW 93 for households with female
labor. Although the program does not have a significant impact on participation in nonagricultural wage
labor, it does have a significant impact on earnings from this kind of work, both overall (ZMW 471) and
for households with female labor (ZMW 154). This significant impact stems from a small (less than 1
percentage point) increase in permanent nonagricultural wage employment for females.
58
Table 12.11: Participation in Agricultural and Nonagricultural Wage Labor, HH Level
Program
Impact
Follow-
up
Program
Impact
Follow-
up
Program
Impact
Follow-
up
All Males Females
Participation in paid
agricultural labor
-0.145
(-3.85)
0.34 -0.0807
(-2.23)
0.26 -0.174
(-4.55)
0.29
Participation in paid non-
agricultural labor
0.0371
(1.67)
0.19 0.0398
(1.71)
0.18 0.0316
(1.58)
0.11
Days in paid agriculture
(year)
-13.75
(-2.76)
35.69 -3.036
(-0.73)
22.34 -12.37
(-5.02)
18.64
Days in paid nonagriculture
(year)
3.025
(1.04)
19.93 2.082
(0.80)
15.53 1.088
(0.63)
8.05
Earnings in paid agriculture
(year)
-67.62
(-1.25)
337.04 22.44
(0.46)
221.13 -93.43
(-3.63)
168.16
Earnings in paid
nonagriculture (year)
471.65
(1.97)
693.37 380.60
(1.45)
666.33 153.64
(2.17)
182.40
N 2296 2296 1764 1764 2282 2282
NOTE: Estimations use single difference modeling. Robust t-statistics clustered at the CWAC level are in parentheses. Bold
indicates that they are significant at p < .05. All estimations control for household size, recipient age, education and marital
status, districts, and household demographic composition. Follow-up refers to follow-up mean value of indicator shown in the
preceding column.
If not working in agricultural wage labor, what do the male and female adults in beneficiary households
do with their time? Part of that time is spent working in the family’s nonfarm enterprise—the CGP leads
to a 16 percentage point increase in the share of households that had labor dedicated to nonfarm
enterprise activity, with an average increase of 1.57 days a week in terms of intensity (Table 12.12). The
impact is somewhat higher for female labor (16 percentage points and 0.98 of a day a week in terms of
intensity compared with 12 percentage points and 0.62 of a day a week).
Table 12.12: Participation and Days Worked in Nonfarm Enterprise, HH Level
Program
Impact
Follow-
up
Program
Impact
Follow-
up
Program
Impact
Follow-
up
All Males Females
Participation in NFE 0.171
(4.69)
0.38 0.120
(4.78)
0.18 0.156
(4.58)
0.33
Days worked (last week)
in NFE
1.573
(4.38)
2.64 0.618
(3.57)
0.94 0.984
(4.50)
1.76
N 2202 2202 2102 2102 2197 2197
NOTE: Estimations use single difference modeling. Robust t-statistics clustered at the CWAC level are in parentheses. Bold
indicates that they are significant at p < .05. All estimations control for household size, recipient age, education and marital
status, districts, and household demographic composition. Follow-up refers to follow-up mean value of indicator shown in the
preceding column.
We expect the CGP to lead to an increase in the intensity of labor on the farm, given the productive
impacts described above. Indeed, households with male labor spend an extra 13 days on their own farm
agricultural activities (Table 12.13). Overall, beneficiary households spend an extra 20 days on their own
farm labor (significant at the 10 percent level). Finally, adults may also increase their time in domestic
59
chores or child care or simply leisure, but we did not collect data on these common household activities,
which can all lead to an increase in family well-being.
Table 12.13: Participation and Days Worked on Own Farm Agriculture, HH Level
Program
Impact
Follow-
up
Program
Impact
Follow-
up
Program
Impact
Follow-
up
All Males Females
Participation on own
farm
-0.0133
(-0.61)
0.92 0.0170
(0.71)
0.79 -0.0140
(-0.65)
0.92
Days worked (last year)
on own farm
20.19
(1.84)
145.76 13.27
(2.00)
71.60 8.242
(1.50)
78.45
N 2202 2202 2102 2102 2197 2197
NOTE: Estimations use single difference modeling. Robust t-statistics clustered at the CWAC level are in parentheses. Bold
indicates that they are significant at p < .05. All estimations control for household size, recipient age, education and marital
status, districts, and household demographic composition. Follow-up refers to follow-up mean value of indicator shown in the
preceding column.
Finally, in terms of children, overall the program has not had any impact on child labor, either paid or
unpaid (Table 12.14). Given program impacts on household productive activities and adult labor supply,
along with findings on reducing child labor from cash transfer programs in other countries, these results
suggest the need for further detailed study.
Table 12.14: Impact of CGP on Child Labor Supply (Share), Individual Level
Program
Impact
Baseline Program
Impact
Baseline Program
Impact
Baseline
All Females Males
Total 0.047
(0.05)
0.53 0.016
(0.05)
0.54 0.083
(0.06)
0.52
Paid -0.017
(0.01)
0.05 -0.014
(0.01)
0.06 -0.017
(0.02)
0.04
Unpaid 0.039
(0.05)
0.48 0.001
(0.06)
0.49 0.080
(0.06)
0.48
N 8062 4182 4053 2117 4009 2065
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC
level are in parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient
age, education and marital status, districts, household demographic composition and a vector of cluster-level prices. Follow-up
refers to follow-up mean value of indicator shown in the preceding column.
60
Simulations of the Impact on the Local Economy20
The CGP is likely to have an economic impact beyond that on beneficiary households described above.
As they spend CGP transfers, beneficiary households transmit the impact of the program to other
households inside and outside the communities in which they live.
The structure of the Local Economy-wide Impact Evaluation (LEWIE) model is centered on the principal
economic activities in which beneficiary and nonbeneficiary households participate, the households’
income sources, and the goods and services on which households spend their income. Household
participate in crop and livestock production; in retail, service, and other production activities; and in the
labor market. The retail sector includes village stores, which obtain most of their goods outside the
village, as well as stores in nearby villages, towns, and the rest of Zambia. Production activities use
different factors: hired labor, family labor, land, capital, livestock, and purchased inputs, some of which
are obtained inside the village and some outside.
Households consume and produce local commodities and can export production or import outside
goods. The LEWIE model incorporates both CGP and nearby villages and towns, which may have a large
market visited by people from many communities. This area is called the Zone of Influence (ZOI). Table
12.15 illustrates how expenditures vary across space by commodity. More than 70 percent of all retail
purchases—and 90 percent of livestock—are made in both the CGP or nearby village. These high shares
lay the basis for potential income multipliers within the local economy. Purchased inputs for crop
production and business retail, however, tend to be made at the town or beyond level. The linkages
between the ZOI and the rest of the economy determine how the transfer flows between households in
the local economy and whether spillovers accrue to households locally.
Table 12.15: Locations of Purchases of Different Commodities and Factors (share)
Items Purchased
Village
Nearby
Village
Town
Elsewhere
(incl govt)
Retail items purchased by households 0.545 0.172 0.281 0.002
Purchased input for crop production 0.117 0.095 0.535 0.252
Retail inputs purchased by businesses 0.172 0.095 0.444 0.289
Animal products purchased by
households
0.820
0.131 0.049 0.000
NOTE: Data are taken from CGP impact evaluation household surveys and business enterprise survey.
The LEWIE model simulation shows that the CGP has a potential total income multiplier of ZMW 1.79 in
nominal terms, with a 90 percent confidence interval (CI) of 1.73-1.85. That is, each Kwacha transferred
to poor households can raise the local income by ZMW 1.79 (Table 12.16).
20
This section summarizes results from the local economy simulations of program effects, which are reported fully
in Thome, K., Taylor, J. E., Davis, B., Handa, S., Seidenfeld, D., & Tembo, G. (2013). Local Economy-wide Impact
Evaluation (LEWIE) of Zambia’s Child Grant Program (PtoP project report). Rome, Italy: FAO and the World Bank.
61
Table 12.16: Simulated CGP Income Multiplier
Total Income
Multiplier
Confidence
Interval
Eligible
Households
Ineligible
Households
Nominal 1.79
1.73 to 1.84 1.17 0.62
Real 1.34
1.29 to 1.39 1.05 0.30
NOTE: Estimations are based on the LEWIE model. “Nominal” impacts refer to the
simulation where producers respond perfectly to increases in demand. “Real” allows
for an imperfect supply response, which leads to price inflation in the ZOI.
Eligible households receive the direct benefit of the transfer, and ineligible households receive the bulk
of the indirect benefit. Of the ZMW 1.79 nominal income multiplier, ineligible households receive ZMW
0.62 for each Kwacha given to eligible households; the eligible households receive the value of the
transfer plus an extra ZMW 0.17, for a total of ZMW 1.17 (Table 12.16). Beneficiary households thus
benefit both directly and indirectly from the transfer program.
The impact of the CGP varies considerably across sectors of economic activity. The cash transfers
stimulate the production of crops and livestock by ZMW 0.47 and ZMW 0.09 per Kwacha transferred,
respectively. The largest positive multiplier effects are on retail (ZMW 1.91), which is what we expect
because the bulk of spending on retail is done within the CGP or nearby villages (that is, within the ZOI).
Like the income multipliers, much of these impacts occur in the ineligible households (Table 12.17)
because they own the shops and retail outlets.
Table 12.17: Simulated CGP Production Multiplier
Production
Multiplier
Confidence
Interval
Eligible
Households
Ineligible
Households
Crop production 0.47 0.41 to 0.52 0.18 0.29
Livestock
production
0.09 0.07 to 0.10 0.02 0.07
Services 0.28 0.26 to 0.30 0.10 0.18
Other production 0.02 0.01 to 0.02 0.01 0.01
Retail 1.91 1.81 to 2.01 0.40 1.51
NOTE: Estimations are based on the LEWIE model.
These simulations assume that producers can respond immediately to the increase in demand brought
about by the greater purchasing power of beneficiaries. But if local production or supply of goods do not
(or cannot) increase sufficiently to meet the increased demand brought on by the CGP, prices will
increase. This will raise consumption costs for all households and could result in a real-income multiplier
that is lower than the nominal multiplier. According to the CGP LEWIE, this real income multiplier of the
program could be as low as ZMW 1.34 (CI: 1.29–1.39; bottom row of Table 12.16).
These simulations illustrate that without efforts to ensure an adequate supply response in the local
economy, part of the program’s impact may be inflationary rather than real. Even a relatively small
62
increase in the local current price index (CPI) can result in a smaller real income multiplier because it
potentially affects all expenditures of all household groups. The higher the local supply response, the
larger is the real expansion in the local economy and the smaller the resulting inflation effect. The
analysis of inflation trends presented in Annex 1 suggests that there is no excess price inflation in
intervention areas due to the CGP. It appears that the “nominal” effects reported in Table 8.16 are a
closer approximation of the likely spillover impacts of the CGP in these areas.
63
XIII. Discussion and Conclusion
The design of the impact evaluation of Zambia’s CGP represents a gold standard in evaluation research
in that it involves a large, multisite sample with an experimental control group, a baseline measurement,
and repeated post-intervention measures. Attrition in the first follow-up is reasonable at 10 percent,
with no differential attrition, thus preserving the experimental balance created at baseline.
Consequently, results presented here can be interpreted as causal effects of the CGP on the indicators
reported rather than confounding factors that have not been accounted for.
Operational Performance and Theory of Change
Results from the data collected at the 24-month follow-up on perceptions of program operations
indicate that recipients are, by and large, satisfied with the operation of the program: transfers are
being delivered in a timely manner, and out-of-pocket costs of collecting payments are small. The
timely, predictable delivery of cash and the low cost to beneficiaries of collecting the money are
essential preconditions for ensuring positive program impacts. These preconditions appear to have been
met in the CGP.
The challenge with evaluating the impact of an unconditional cash transfer program is that households
are free to use the money as they see fit. Because cash is fungible, impacts might be found anywhere,
depending on the preferences and constraints of each individual household. We have addressed this
challenge by laying out a theoretical framework for the behavioral response of households and by using
pre-program data to estimate income effects for different indicators to give us an idea of the
preferences of households and how they are likely to use the cash transfer. We use our theory and
these expected impacts to guide our analysis; in addition, deviations between predicted and actual
impacts provide insights about how households change their previous behaviors in response to
participation in the CGP.
Consumption and Food Security
Consistent with the relative poverty of households and our theory, we see large impacts of the CGP on
consumption expenditures, which increase by more than the per capita value of the transfer. This result
for consumption suggests that the beneficiary households use the transfer to produce more, and this
can be attributed to increased production (see section 12). Also consistent with ex-ante behavior, three-
fourths of the transfer is spent on food. However, contrary to expectation, a significant portion of the
transfer is spent on clothing (6.1 percent) and health (7.1 percent); this result is consistent with what
beneficiaries report to be their obligation and responsibility as program participants, namely, food,
clothing, and health of young children. This behavioral response is interesting given that the program is
unconditional and no punitive sanctions are associated with patterns of spending. Within food spending,
we find that the CGP significantly increases both caloric intake (cereals) and protein (meats, dairy). In
addition, we see significant impacts on sugars, oils and fats, and pulses, which is somewhat contrary to
the ex-ante predictions. These differences, particularly for oils and fats, contribute to diet diversity and
likely imply more cooking at home. The overall increase in total and food spending is borne out by
significant improvements in food security, measured both through the FANTA food security index and
self-reports by households of the number of meals eaten per day, as well as their perceptions of
64
whether their life has improved over the previous year. Together, this is extremely strong evidence of
positive impacts on the overall monetary well-being of CGP participants.
Individual Impacts on Children
Impacts on specific indicators for individuals within the household are necessarily second-round effects
in that they work their way through the spending and time-use allocations brought about by the cash
transfer. Among young children, we find significant program impacts on reducing diarrhea (5 percentage
points), increasing IYCF for children ages 6 to 24 months (22 percentage points), and reducing wasting
among this same group. Among all children ages 0 to 60 months, we find some signs that their weight is
improving, although impacts are just outside the level of statistical significance. There are definite
indications that the caring environment for young children ages 3 to 7 has improved as a result of the
CGP, with children receiving significantly more support for learning (13 percentage points) and learning
materials (1 percentage point, or a 68 percent increase over baseline). Most of these impacts are
consistent with what we predicted at baseline and are also consistent with the expenditure impacts that
show increases in calorie and protein availability, clothing, and health spending.
Among children ages 6 to 17, our ex-ante analysis suggested that the CGP would improve their material
welfare but not schooling or health, and this is what we find at follow-up. The CGP has had a large
impact (33 percentage point increase) on ensuring that children’s material needs are met (possession of
blanket, shoes, clothing), with the overall effect particularly driven by shoes. This is exactly consistent
with the expenditure results, which show a large impact on clothing expenditure.
Maternal Health and Women’s Status
This report also presents findings on a set of maternal health and women’s status indicators, although
we did not investigate these at baseline to predict impacts, and they are not featured prominently in the
conceptual framework because the link between income and these indicators is not well-established
given the importance of supply-side factors. Indeed, the evidence reported here shows no impacts on
maternal health indicators, such as antenatal care, assisted delivery, morbidity, or chronic illness. We do
not find evidence of impacts on women’s decision making within the household. However, we do see a
positive impact on the propensity for women to save money (20 percentage points) in the reference
period, the amount saved, and their expectation that their lives will be better in 3 years (12 percentage
points). We also find significant program impacts on self-reported general health status among women,
which, taken together with their expectations about the future and their ability to save, may reflect an
overall optimism about their lives.
Economic Impacts
The theoretical framework posits that once basic consumption needs are satisfied households may
begin using the cash transfer to bolster their livelihoods strategies, either by diversifying income sources
or expanding their current productive activity. In the case of the CGP, we observe both types of
economic impacts. Over 80 percent of CGP households are engaged in agricultural production, and we
report positive impacts of the program on both investment in production and the value of that
production. For example, there is a program impact of 18 percentage points in the share of households
65
with any expenditure on productive inputs (seeds, fertilizer, labor), and a 50 percent increase in the
overall value of agricultural commodities harvested.
The CGP also leads to income diversification among recipients. There is a significant positive impact on
the share of households owning livestock (21 percentage points) from a base of only 49 percent at
baseline. The CGP also boosted non-farm economic activity, with an impact of 17 percentage points in
the proportion of households engaged in any nonfarm enterprise and corresponding increases in
business asset ownership, months of operation, revenue, and profit.
As to be expected given the impacts on both existing economic activity and new activity, the CGP leads
to interesting and generally welfare-enhancing patterns of labor reallocation. Households in the
program have reduced their engagement in casual labor (14 fewer days), typically the least productive
form of work in rural settings, and increased their time to own-farm activity (20 more days). In addition,
more household time is devoted to nonfarm enterprises, which is usually the most economically
productive type of work in these settings. In terms of demographic differences, women tend to increase
their time in nonfarm business, whereas men increase their time for working on their own farm. The
program has no discernible impacts on child labor.
Local Economy Effects
A study being conducted in parallel to the main evaluation seeks to measure the impact of the CGP on
the local economy, for both beneficiaries and nonbeneficiaries. Simulations show that the CGP has a
potential total income multiplier of ZMW 1.79—that is, each Kwacha transferred to poor households can
raise local income by ZMW 1.79. Beneficiary households receive the direct benefit of the transfer,
whereas ineligible households receive the bulk of the indirect benefit. Of the ZMW 1.79 nominal income
multiplier, ineligible households receive ZMW 0.62 for each Kwacha given to beneficiary households,
while the beneficiary households receive the value of the transfer plus an extra ZMW 0.17, for a total of
ZMW 1.17. Beneficiary households thus benefit both directly and indirectly from the transfer program.
More important, the CGP can also have a significant impact on the incomes of nonbeneficiaries, a fact
that has not been fully realized or documented in other cash transfer evaluations.
Conclusions
The CGP has generated positive impacts on a range of indicators identified in the conceptual framework
as being plausible. What is particularly exciting about the results presented here is that the CGP not only
addresses the immediate consumption and food security needs of recipients but also leads to significant
increases in the productive capacity of households, both by supporting the expansion of existing
economic activity and by enabling their diversification into new activity. There is also evidence that the
program is beginning to have an impact on young children, with improved feeding and reduced wasting
among children ages 6 to 24 months, reduced morbidity among children ages 0 to 60 months, and
improvements in weight among all children ages 0 to 60 months (although not statistically significant).
The learning and developmental environment for children ages 3 to 7 has also significantly improved, as
has access to basic needs (clothing, shoes, blanket) among children ages 5 to 17. Table 13.1 links each
program objective with the indicators reported here.
66
Table 13.1: Summary of Impacts in Areas Directly Linked to CGP Objectives
Supplement and not replace household income
Increase of ZMW 15 in monthly per capita consumption
expenditure
Reduction of 11 percentage points in poverty gap and
squared poverty gap
Increase the number of households having a second
meal per day
Increase of 8 percentage points in households with 2+
meals per day
Increase of 22 percentage points in proportion of
children ages 6 to 24 months receiving minimum
feeding requirements
Reduce the rate of mortality and morbidity of children
under 5
Reduction in diarrhea of 5 percentage points
Reduce stunting and wasting among children under 5
Increase in weight-for-height of 0.196 z-scores among
children ages 3 to 5 years
Increase in weight-for-weight and weight-for-age of
0.118 and 0.128, respectively, among children ages 0
to 5, but no statistically significant effects
Increase the number of children enrolled in and
attending primary school
No statistically significant effects
Increase the number of households owning assets such
as livestock
Increase of 21 percentage points in households owning
any livestock
Increase of 4.5 percentage points in households owning
any nonfarm business assets
67
Annex 1: Prices in the CGP Evaluation Sample
There is a concern that in the remote villages of Zambia where the CGP operates, a large influx of cash
to the community may lead to inflation if supply cannot adequately respond to the new increase in
demand for goods and services. We implemented a community questionnaire as part of the survey
fieldwork in which we collected prices on 12 key consumption items. We inflated the reported values in
2010 to 2012 units using the all-Zambia CPI and checked to see whether there was any excess inflation
in intervention communities relative to control communities, a sign of supply bottlenecks that might
cause inflationary pressure with the existence of the program.
Table A1.1 begins by simply comparing prices for each item across time among all communities. Column
3, which reports t-statistics for mean differences, shows no excess inflation in these communities once
we account for the all-Zambia CPI. If anything, in a few cases (e.g., cooking oil, sugar), prices are
somewhat lower in the evaluation communities in 2012 than in 2010.
Table A1.1: Community Prices Over Time (in Zambian Kwacha)
Baseline 24-Month
Follow-up
t-statistic
Maize grain price 30.58 25.99 0.02
Rice price 5.31 4.75 0.22
Bean price 7.28 11.24 -1.74
Dry fish price 4.11 4.48 0.71
Chicken price 17.24 16.70 -0.88
Cooking oil price 13.12 11.75 -2.04
Sugar price 9.21 9.28 -1.94
Table salt price 8.12 5.16 -0.66
Toilet soap price 6.92 5.53 -0.39
Laundry soap price 6.76 6.25 0.99
Panadol price 4.50 5.07 0.66
Secondary school fee 402.64 712.93 -0.36
N 90 90
NOTE: t-tests estimates provided. Baseline prices inflated to 2012 levels.
Table A1.2 reports difference-in-difference estimates that effectively compare the change in a price over
this period between treatment and control households in a manner similar to program impact estimates
reported in the main text. We are interested in whether the existence of the program has led to an
increase in a price relative to control communities; we find no evidence of excess inflation in treatment
communities. Indeed, the only statistically significant impact is for cooking oil and that shows a relative
decline in price rather than an increase.
68
Table A1.2 Community Prices, by CGP Treatment
Program
Impact
Baseline 24-Month
Treatment
24-Month
Control
Maize grain price -4.30 30.58 26.01 25.98
(-1.44)
Rice price 0.17 5.31 4.81 4.69
(0.24)
Bean price -2.48 7.28 10.16 12.33
(-1.73)
Dry fish price 0.62 4.11 4.72 4.24
(0.58)
Chicken price 1.17 17.24 15.93 17.48
(0.55)
Cooking oil price -2.01 13.12 11.04 12.46
(-2.54)
Sugar price -1.14 9.21 8.87 9.70
(-1.90)
Table salt price -1.00 8.12 5.00 5.32
(-0.58)
Toilet soap price 0.02 6.92 5.46 5.60
(0.03)
Laundry soap price 0.08 6.76 6.44 6.06
(0.13)
Panadol price 1.02 4.50 7.55 6.53
(0.63)
Secondary school fee -102.68 402.64 689.29 736.56
(-0.73)
N 180 90 45 45
NOTE: Estimations use difference-in-difference modeling. Cluster robust t-statistics are in parentheses.
Bold indicates that they are significant at p < .05. All estimations control for District. Baseline prices are
inflated to 2012 levels.
69
Annex 2: Difference-in-Differences Estimation
The statistical approach we take to derive average treatment effects of the CGP is the difference-in-
differences (DD) estimator. This entails calculating the change in an indicator (Y), such as food
consumption, between baseline and follow-up period for treatment and comparison group units and
comparing the magnitude of these changes. Figure A2.1 illustrates how the estimate of differences in
differences between treatment (T) and control (C) groups is computed. The top row shows the baseline
and postintervention values of the indicator (Y), and the last cell in that row depicts the change or
difference in the value of the outcome for T units. The second row shows the value of the indicator at
baseline and postintervention for comparison group units, and the last cell illustrates the change or
difference in the value of this indicator over time. The difference between these two differences
(treatment vs. control), shown in the shaded cell in Figure A2.1, is the difference-in-differences or
double-difference estimator.
Figure A2.1: The Difference-in-Differences (DD) Estimator
Baseline (2010) Post (2012) 1st difference
Treatment (T) YT
2010 YT
2012 ΔYT=(Y
T2012-Y
T2010)
Comparison (C) YC
2010 YC
2012 ΔYC=(Y
C2012-Y
C2010)
Difference in differences DD =
(ΔYT – ΔY
C)
The DD is one of the strongest estimators available in the evaluation literature (Shadish et al., 2002).
Two key features of this design are particularly attractive for deriving unbiased program impacts. First,
using pre- and posttreatment measures allows us to “difference” out unmeasured fixed (i.e., time-
invariant) family or individual characteristics that may affect outcomes, such as motivation, health
endowment, mental capacity, and unobserved productivity. It also allows us to benchmark the change in
the indicator against its value in the absence of treatment. Second, using the change in a control group
as a comparison allows us to account for general trends in the value of the outcome. For example, if
there is a general increase in school enrollment owing to expansion of school access, deriving treatment
effects based only on the treatment group will confound program impacts on schooling with the general
trend increase in schooling.
The key assumption underpinning the DD is that there is no systematic unobserved time-varying
difference between the T and C groups. For example, if the T group changes its preference for schooling
over time but the C group does not, then we would attribute a greater increase in schooling in T to the
program rather than to this unobserved time-varying change in characteristic. In practice, the random
assignment to T and C, the geographical proximity of the samples, and the rather short duration
between pre- and postintervention measurements will make this assumption quite reasonable.
When treatment and comparison units are selected randomly and their characteristics are perfectly
balanced, the simple mean differences as shown in Figure A2.1 are usually sufficient to derive unbiased
estimates of program impact. However in large-scale social experiments, it is typical to estimate the DD
70
in a multivariate framework, controlling for other potential intervening factors that might not be
perfectly balanced across T and C units and/or are strong predictors of the outcome (Y). Not only does
this allow us to control for possible confounders, it also increases the efficiency of our estimates by
reducing the residual variance in the model. Of course, there is an important weakness to the
multivariate approach, which is that overfitting the statistical model can wash-away program effects
that work through the control variables. For example, if we control for the number of young children in
the household when estimating treatment effects on nutrition, and if the program improves nutrition
through decreases in fertility (through the well-known child quantity-quality trade-off), then we may not
estimate a positive treatment effect when controlling for the number of young children, even though
the program actually has an impact on nutrition.
Cross-Section Analysis of Selected Indicators
One data issue distinguishes the nonagricultural enterprise and labor analysis from the analysis used in
the rest of the report. Both a detailed labor module and a nonagricultural enterprise model were
included in the 2012 follow-up questionnaire but not in 2010. Consequently, we have only one
observation per household and per individual for most of the labor and nonagricultural enterprise
outcomes of interest. Impact estimates for these indicators are derived using multivariate cross-section
analyses. We also experimented with inverse probability weight estimators but these yielded similar
results given the excellent balancing properties at baseline.
71
Annex 3: Mean Differences at Baseline for Attrition Analysis
Table A3.1: Household-Level Control Comparisons (Control Versus Treatment for Respondent Households)
Variables Control N1 Treatment N2
Mean
Difference p-value
Household size 5.66 1167 5.74 1128 -0.08 0.6489
Number of people ages 0-5 1.91 1167 1.89 1128 0.02 0.7244
Distance to food market 21.29 777 15.29 754 6.00 0.3245
Distance to health facility 9.48 1061 9.29 1024 0.18 0.8986
Yes/no whether household was affected by drought 0.05 1167 0.04 1128 0.01 0.7416
Yes/no whether household was affected by flood 0.06 1167 0.03 1128 0.03 0.2007
Yes/no whether household was affected by any shocks 0.17 1167 0.16 1128 0.01 0.8431
T-tests clustered on the CWAC level.
Table A3.2: Household-Level Outcome Comparisons (Control Versus Treatment for Respondent Households)
Variables Control N1 Treatment N2
Mean
Difference p-value
Per capita food expenditure, kwacha 29.20 1167 30.86 1128 -1.66 0.4675
Food share of total household expenditure 0.71 1167 0.72 1127 0.00 0.7885
Cereal as share of total food expenditure 0.32 1167 0.35 1125 -0.03 0.3583
Roots and tubers as share of total food expenditure 0.16 1167 0.14 1125 0.02 0.5568
Pulses and legumes as share of total food expenditure 0.03 1167 0.03 1125 0.00 0.5668
Fruits and vegetables as share of total food
expenditure 0.23 1167 0.21 1125 0.02 0.2151
Meats, poultry, fish as share of total food expenditure 0.17 1167 0.18 1125 -0.01 0.3718
Total household expenditure per person in the
household 39.65 1167 41.52 1128 -1.87 0.4861
Food security scale 15.31 1150 14.90 1106 0.41 0.4967
T-tests clustered on the CWAC level.
Table A3.3: Children Under 5 Control Comparisons (Control Versus Treatment for Respondent
Households)
Variables Control N1 Treatment N2
Mean
Difference p-value
Person’s age in months 27.18 1983 26.55 1908 0.63
0.1856
Gender 0.51 2225 0.48 2128 0.03 0.0416
Highest grade level the primary caregiver
completed
3.46 2345 3.95 2229 -0.49 0.1000
Received BCG vaccine 0.96 1958 0.97 1878 -0.01
0.1895
Received oral polio vaccine (OPV) 0.96 1956 0.95 1873 0.00
0.7582
Received DPT vaccine 0.95 1953 0.95 1867 0.00
0.8333
T-tests clustered on the CWAC level.
72
Table A3.4: Children Under 5 Outcome Comparisons (Control Versus Treatment for Respondent
Households)
Variables Control N1 Treatment N2
Mean
Difference p-value
Weight in kilograms 11.80 1902 11.86 1823 -0.06 0.8671
Height in centimeters 80.15 1796 79.16 1712 0.99 0.3574
Received vitamin A dose in the 6 months prior to
survey 0.76 1664 0.80 1612 -0.04 0.1841
Had diarrhea in the 2 weeks prior to survey 0.17 1959 0.20 1874 -0.03 0.2486
Had a fever in the 2 weeks prior to survey 0.23 1968 0.23 1888 0.00 0.9114
T-tests clustered on the CWAC level.
Table A3.5: Children Under 5 Anthropometrics (Control Versus Treatment for Respondent Households)
Variables Control N1 Treatment N2
Mean
Difference p-value
Length/height-for-age z-score -1.43 1642 -1.42 1535 -0.01 0.9175
Weight-for-age z-score -0.91 1766 -0.94 1657 0.03 0.6787
Weight-for-length/height z-score -0.15 1641 -0.21 1524 0.06 0.3244
T-tests clustered on the CWAC level.
Table A3.6: Children Ages 3–7 Development Scores (Control Versus Treatment for Respondent Households)
Variables Control N1 Treatment N2
Mean
Difference p-value
Development scale 1 - Played with items 1.50 1427 1.55 1356 -0.05 0.4099
Care scale - Family engagement activities 2.52 1427 2.41 1356 0.11 0.5446
Development scale 2 - Various skills/behaviors 4.39 1427 4.47 1356 -0.08 0.6380
T-tests clustered on the CWAC level.
Table A3.7: Older Child (5–17) Characteristics (Control Versus Treatment for Respondent Households)
Variables Control N1 Treatment N2
Mean
Difference p-value
Female 0.50 2112 0.52 2093 -0.02 0.2934
Maternal orphan 0.08 2112 0.09 2093 -0.01 0.6417
Paternal orphan 0.16 2112 0.18 2093 -0.02 0.3783
OVC 0.21 2112 0.23 2093 -0.02 0.4088
Minimum needs met 0.80 2112 0.77 2093 0.04 0.3390
Ever enrolled in school 0.73 2104 0.72 2085 0.00 0.8308
Currently enrolled in school 0.64 2104 0.64 2085 0.00 0.8783
Full attendance in prior week 0.78 1308 0.80 1278 -0.01 0.6604
Paid or unpaid work 0.59 2081 0.57 2048 0.02 0.5776
Unpaid hours in last 2 weeks 21.23 1213 23.70 1129 -2.47 0.4030
T-tests clustered on the CWAC level.
73
Table A3.9: Household-Level Outcome Comparisons (Full Sample Versus Remaining Sample at 24-Month
Follow-Up)
Variables Full Sample N1
Remaining
Sample N2
Mean
Difference p-value
Per capita food expenditure, kwacha 30.03 2519 30.02 2298 0.01 0.92
Food share of total household expenditure 0.72 2517 0.72 2297 0.00 0.00
Cereal as share of total food expenditure 0.33 2515 0.34 2295 -0.01 <.0001
Roots and tubers as share of total food expenditure 0.17 2519 0.15 2295 0.02 <.0001
Pulses and legumes as share of total food expenditure 0.03 2519 0.03 2295 0.00 0.39
Fruits and vegetables as share of total food
expenditure
0.16 2515 0.22 2295 -0.06 0.78
Meats, poultry, fish as share of total food expenditure 0.18 2515 0.18 2295 0.00 0.35
Total household expenditure per person in the
household
40.43 2515 40.57 2298 -0.14 0.52
Food security scale 15.15 2474 15.1 2259 0.05 0.29
T-tests clustered on the CWAC level.
Table A3.10: Children Under 5 Control Comparisons (Full Sample Versus Sample Remaining at 24-Month
Follow-Up)
Variables Full Sample N1
Remaining
Sample N2
Mean
Difference p-value
Person’s age in months 28.54 224 28.88 168 -0.34 0.0116
Gender 11.79 369 11.83 3729 -0.04 0.7521
Highest grade level the primary caregiver completed 0.72 1420 0.58 1313 0.14 0.4527
Received BCG vaccine 0.95 217 0.94 166 0.01 0.1807
Received Oral Polio Vaccine (OPV) 0.95 217 0.96 166 -0.01 0.8242
Received DPT vaccine 0.94 216 0.94 166 0.00 0.5457
T-tests clustered on the CWAC level.
Table A3.11: Children Under 5 Outcome Comparisons at Baseline (Full Sample Versus Sample
Remaining at 24-Month Follow-Up)
Variables Full Sample N1 Panel N2
Mean
Difference p-value
Weight in kilograms 79.67 352 79.67 3512 0.00 0.9973
Height in centimeters 0.94 390 0.94 3820 0.01 0.6575
Does child have a Health Card? 0.71 328 0.78 3280 -0.07 0.0489
Received vitamin A dose in the
6 months prior to survey 0.20 384 0.19 3837 0.02 0.4805
Had diarrhea in the 2 weeks
prior to survey 0.22 388 0.23 3860 -0.02 0.4852
T-tests clustered on the CWAC level.
74
Table A3.12: Children Under 5 Anthropometrics (Full Sample Versus Sample Remaining at 24-Month
Follow-Up)
Variables
Full
Sample N1 Panel N2
Mean
Difference p-value
Length/height-for-age z-score -1.48 308 -1.43 3181 -0.05 0.5570
Weight-for-age z-score -1.01 338 -0.93 3427 -0.09 0.3789
Weight-for-length/height z-score -0.10 307 -0.18 3169 0.08 0.4176
T-tests clustered on the CWAC level.
Table A3.13: Children (3–7) Development Scores (Full Sample Versus Sample Remaining at 24-Month
Follow-Up)
Variables Full Sample N1 Panel N2
Mean
Difference p-value
Development scale 1 - Played with items 1.41 287 1.52 2787 -0.11 0.1286
Care scale - Family engagement activities 2.21 287 2.47 2787 -0.25 0.1115
Development scale 2 - Various skills/behaviors 4.55 287 4.42 2787 0.13 0.4005
T-tests clustered on the CWAC level.
Table A3.14: Older Child (5–17) Characteristics at Baseline, Full Sample Versus Panel
Variables Full Sample N1 Panel N2
Mean
Differenc
e p-value
Female 0.47 386 0.51 4205 -0.04 0.1024
Maternal orphan 0.10 386 0.08 4205 0.02 0.5071
Paternal orphan 0.20 386 0.17 4205 0.03 0.3422
OVC 0.26 386 0.22 4205 0.03 0.4014
Minimum needs met 0.77 386 0.79 4205 -0.02 0.6211
Ever enrolled in school 0.77 382 0.72 4189 0.05 0.0370
Currently enrolled in school 0.68 382 0.64 4189 0.04 0.1361
Full attendance in prior week 0.80 252 0.79 2586 0.01 0.8481
Paid or unpaid work 0.53 375 0.58 4129 -0.05 0.2702
Unpaid hours last 2 weeks 20.90 195 22.42 2342 -1.52 0.5298
T-tests clustered on the CWAC level.
75
Annex 4: Additional Results on Economics Impacts
Table A4.1: Impact of CGP on Crop Input Use (share) Program Impact Baseline Program Impact Baseline Program Impact Baseline
All HH Size<6 HH Size>5
Total crop exp 0.177
(4.31)
0.23 0.223
(4.52)
0.22 0.134
(2.98)
0.24
Exp seed 0.100
(3.11)
0.13 0.135
(3.60)
0.12 0.067
(1.78)
0.14
Exp hired labor 0.054
(3.69)
0.03 0.072
(3.97)
0.03 0.038
(1.84)
0.04
Exp pesticides 0.002
(0.82)
0.00 0.004
(1.17)
0.00 0.001
(0.39)
0.00
Exp fertilizer 0.032
(2.11)
0.01 0.034
(2.69)
0.01 0.029
(1.35)
0.01
Other crop exp 0.151
(4.00)
0.11 0.153
(3.19)
0.11 0.150
(3.80)
0.11
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC level are in
parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient age, education and marital
status, districts, household demographic composition and a vector of cluster-level prices. Baseline refers to baseline mean value of indicator
shown in the preceding column.
Table A4.2: Impact of CGP on Agricultural Implements (share) Program Impact Baseline Program Impact Baseline Program Impact Baseline
All HH Size<6 HH Size>5
Axe 0.008
(0.22)
0.78 0.005
(0.10)
0.73 0.007
(0.17)
0.81
Pick 0.010
(0.69)
0.03 0.001
(0.05)
0.03 0.019
(1.22)
0.03
Hoe 0.010
(0.56)
0.92 0.002
(0.09)
0.90 0.020
(0.87)
0.93
Hammer 0.044
(3.20)
0.05 0.025
(1.63)
0.04 0.065
(3.15)
0.06
Shovel 0.031
(2.15)
0.06 0.017
(1.09)
0.04 0.044
(1.84)
0.08
Plough 0.036
(1.97)
0.07 0.025
(1.28)
0.05 0.051
(2.10)
0.08
N 4596 2298 2336 1168 2260 1130
NOTE: Estimations use difference-in-difference modeling among panel households. Robust t-statistics clustered at the CWAC level are in
parentheses. Bold indicates that they are significant at p < .05. All estimations control for household size, recipient age, education and marital
status, districts, household demographic composition and a vector of cluster-level prices. Baseline refers to baseline mean value of indicator
shown in the preceding column.
76
Annex 5: The Local Economy-wide Impact Evaluation Model for the CGP
The Local Economy-wide Impact Evaluation (LEWIE) model for the CGP begins by nesting household
farm models for eligible and ineligible households within a region of interest. The household models
describe each group’s production activities, income sources, and expenditure patterns. In a typical
model, households participate in activities such as crop and livestock production, retail, service
provision, and other activities, as well as in the labor market. These activities, as well as household
expenditures, are modeled using data from household surveys.
Household groups in a given village are linked through local trade, and villages are linked through
regional trade. The entire program region interacts with the rest of the country, importing and exporting
goods and selling labor. Interactions among households within the program area and between the
program area and the rest of the economy are modeled using the survey data. The parameters in the
LEWIE model are estimated econometrically. Sensitivity analysis, combined with Monte Carlo methods,
allows testing the robustness of simulated impacts for errors in parameter estimates and model
assumptions.
The LEWIE model is built for treatment and control villages and includes households both eligible and
ineligible for inclusion in the CGP. The Zambia CGP LEWIE draws on baseline and follow-up data
collected in 2010 and 2012 in the three program districts for the randomized controlled trial impact
evaluation of the CGP. The LEWIE model also used the business enterprise survey that was implemented
at follow-up, as well as the nationally representative Living Conditions Measurement Survey (LCMS).
The simulations assume that locally grown crops, livestock, retail, and other services, including labor, are
traded locally. Given high transaction costs with the rest of the country and abroad, it is reasonable to
assume that the prices of the goods produced are determined in local markets. A nearly perfectly elastic
labor supply (η=100) is assumed, which reflects excess labor supply in rural Zambia. This can be
expected to lower inflationary pressures from the program by limiting wage increases. It does not
remove inflationary pressures completely, however, because land and capital constraints may continue
to limit the local supply response. More detail on the methodology, as well as the complete results,
including robustness checks.21
21 Thome, K., Taylor, J. E., Handa, S., Seidenfeld, D., Tembo, G., & Davis, B. (2013). Local Economy-wide Impact
Evaluation (LEWIE) of Zambia’s Child Grant Program (PtoP project report). Rome, Italy: FAO and the World Bank.
77
Annex 6: Community Profile
Although the CGP provides cash directly to households, the cash can also have an impact on the broader
community. The program injects cash into local businesses and other households in the community
through direct spending on locally sourced goods and services by beneficiary households, thus
increasing the local demand on these items. These changes in demand could alter prices, wages, and
availability of credit if the market is unable to respond to that demand. In addition to these potential
effects on the community, the CGP might also improve the functioning of local governance committees
used to implement the program.
The CGP study includes an investigation of the program’s impacts on community dynamics, including
economic activities, with a particular focus on child labor, access to credit, and governance. We find no
impacts of the program on child labor at the community level; however, the amount of child labor
reported has declined over time in both the treatment and control communities. The CGP does not
affect prices or wages, suggesting that the market can meet the new demand. Additionally, the CGP
expands the ability of beneficiary households to secure credit. Last, we find that the CGP improves some
elements of governance in the local structures through which it operates.
Community Profile Study Design
Palm Associates, under the direction of AIR, implemented the data collection for the CGP surveys,
including the community survey. The research team administered a community survey in each of the 90
Community Welfare Assistance Committees (CWACs) across the three program districts: Kaputa, Kalabo,
and Shangombo. As described in the methodology section, the CWACs were randomly assigned to
intervention or control group with 45 communities in each group.22 The community survey was collected
in conjunction with the 24-month follow-up household data collection in September and October 2013.
To investigate community level impacts, a pair of enumerators conducted interviews with a group of key
informants. On average 13.6 informants (36 percent female) participated in each interview, including
the village head, Area Coordinating Committee (ACC) members, CWAC members, government officials,
and NGO workers. In this section we compare changes over time from baseline to follow-up between
these communities by treatment status using difference in difference analysis. We include district fixed
effects to account for clustering of CWACs within a district.
Description of the Communities
To provide a context for understanding the communities, this section describes the population of the
communities, the availability of key facilities to these communities, and shocks experienced by them.
These are poor rural communities, located far from urban centers and associated markets, facilities, and
resources. At follow-up, the communities reported a median population of 1,525 people, with a median
of 355 households. More households have moved into both the intervention and control communities
than have moved out of these communities. On average, 19.9 households have moved into each
22
In the baseline report, the findings were based on 80 community surveys. With further data cleaning, 4
additional community surveys were included in the baseline findings described in this report.
78
community during the past 2 years and 12.9 households have moved out.23 There is no evidence that
the CGP has an impact on household movement because there are no differences in migration between
intervention and control communities.
Increased use of health and education facilities represents one goal of the CGP, therefore we investigate
the availability of these facilities in the community. Ninety percent of CWACs have a primary school,
with 70 percent of primary schools government owned and 28 percent community owned. For the
CWACs without a local primary school, the distance may represent a barrier to access to primary
schooling, especially for younger children and girls. Only three CWACs have a secondary school,
indicating limited local opportunities for secondary school. A total of 31 health facilities serve the three
rural districts of Zambia included in this study. Of these facilities 39 percent are health centers, 39
percent are health posts, and 13 percent are dispensaries (two facilities are not classified). The
household survey reveals a person walks an average of 9 km to reach a health facility, indicating
distances could be a challenge in accessing care, especially when ill.
We look at the various shocks experienced by the community to better understand exogenous factors
that affect their well-being and economic situation. These include positive shocks, such as the opening
of roads that improve accessibility and potentially open trading markets, and negative shocks, such as
floods or droughts that destroy property or food sources. We find no differences between the
intervention groups’ and control groups’ experiences of positive or negative shocks. This helps eliminate
alternative explanations for observed impacts at household and individual levels. This equivalence is also
another signal that randomization has worked. Shocks are moderators on the impact of the cash
transfer, making them weaker or stronger depending on local conditions in the community. In the
analysis of household data, shocks are used as control variables. Table A6.1 shows the various shocks
experienced in the communities. Just under half (45 percent) of the communities experienced any
positive shock, whereas all communities experienced at least one bad shock.
Table A6.1: Differences in Shocks Experienced at 24 month Follow-Up, by Control and Intervention
Control Intervention Mean
difference
t-statistic
Good External Shocks
School constructed 0.13 0.16 0.02 0.2966
Road constructed 0.04 0.07 0.02 0.4556
Health facility constructed 0.02 0.02 0.00 0.0000
New employment opportunity available 0.07 0.13 0.07 1.0488
Development projected started 0.38 0.42 0.04 0.4260
Bad External Shocks
Massive job lay-offs 0.09 0.07 -0.02 -0.3895
Sharp changes in prices 0.91 0.80 -0.11 -1.5014
Human disease/epidemic 0.78 0.84 0.07 0.8018
Livestock disease 0.80 0.82 0.02 0.2664
23
t(86) = 2.690, p < .05
79
Crop disease 0.67 0.64 -0.02 -0.2194
Flood 0.58 0.60 0.02 0.2119
Drought 0.58 0.62 -0.04 -0.4260
N 45 45
NOTE: t-statistics are provided. Bold indicates that they are significant at p < .05.
Economic Activity
We investigate the impact of the program on economic activity across the entire community
(beneficiaries and nonbeneficiaries of the program living in treatment communities). The program does
not have an impact on child labor; however, the amount of child labor has declined over the 2-year
period in both the intervention and control communities. The decline in child labor includes the number
of communities that have children who work as well as the proportion of children who work within
those communities. On average, 78 percent of the communities have children under the age of 16 who
work for money compared with 92 percent at baseline.24 Further, the proportion of children working in
the village has also declined since baseline. Of the villages where children worked at baseline, 64
percent reported that more than half the children participated in some sort of work for money, whereas
only 39 percent of communities report the same 2 years later.25 This could reflect the economic
improvements in Zambia and could perhaps reflect the success of social marketing campaigns focused
on reducing child labor and promoting education. We also see changes over time in the type of
livelihood activities done by children. The majority of children who work (89 percent) engage in
domestic work or farming as the primary form of labor. This is a shift from baseline where only 57
percent of children worked in domestic and home farm labor activities and were more engaged in
fishing, trading, and industry-related work.26 On average, when children are paid, they earn ZMW 8.60
daily, similar to the pay of adults with no difference between groups or over time.
The CGP could open new opportunities for livelihood and change participation in the various types of
labor market. However, in looking at the profile of livelihood activities for adults, we see no changes to
the types of activities in which these communities engage. Similar to baseline, crop farming is the
primary economic activity in 89 percent of the villages. Key secondary livelihood activities include fishing
(37 percent), trade/business (19 percent), and farming livestock (14 percent).
One concern about adding cash to these poor rural communities is the potential for an inflationary
effect on prices and wages, especially if supply cannot adequately respond to the new increase in
demand for goods and services. As described in Annex 1: Prices in the CGP Evaluation Sample, there has
been no excessive inflation in the intervention communities compared with the control communities.
The CGP does not cause an increase in prices in these communities. Similarly, there is no real economic
impact on adult wages. The mean daily wage for men is ZMW 12.52 (approximately $2.50 USD daily),
and women is ZMW 9.78 (approximately $2 USD daily). There are no differences between control and
intervention groups.
24
t(82) = 2.528, p < .05 25
t(58) = 3.231, p < .05 26
t(55) = 4.180, p < .05
80
Access to Credit
Borrowing is a short-term way to alleviate financial shortfalls often associated with poverty. Borrowing
helps households cope with emergencies, smooth out consumption, or even seize small investment
opportunities that could improve their lives. We find that the CGP improves community access to credit.
When someone needs money in times of emergency or to make a large purchase, such as fertilizer,
intervention communities have a greater likelihood of gaining access to credit compared with control
communities. At baseline, only 12 percent of the communities had someone from whom to borrow
money. At follow-up, 40 percent of the intervention communities have a lender, whereas only 9 percent
of the control communities do. The impact of the CGP program is significant, explaining 33 percentage
points of the difference. Given the small sample sizes for subsequent questions about borrowing, we can
only describe the borrowing profile. At baseline, 22 percent of borrowers had to provide collateral to
receive a loan; at follow-up, no one has to provide collateral. The typical median loan is ZMW 100. The
CGP helps open access to additional resources when a household requires them.
Governance
The CGP uses government structures at the community and area levels to manage and administer the
program in rural areas. By engaging in these local structures, the CGP could improve how they function.
These changes could be in how the committees are structured, how often they meet, and how decisions
are made. We find that the CGP has an impact on committee composition at the area level, but not at
community level. We also find that the CGP improves the frequency with which committees meet.
However, there are no program impacts on participation in local decision making (Table A6.2).
Two key committees enable community participation in government: the Area Coordinating Committee
(ACC) and the CWAC. The ACC is a subdistrict structure covering 8 to 12 CWACs. The CWAC is a
community-level structure, covering from as few as 20 up to approximately 500 households. While these
are existing structures, the SCT program leverages these committees to ensure the smooth functioning
of the program. The ACC comprises members from the respective CWACs and is responsible for verifying
potential and actual beneficiaries, monitoring the performance of the CWACs, and handling grievances.
The CWACs comprise members from the community and are responsible for raising awareness about
the SCTs, identifying beneficiary households, communicating details about payments with households,
and counseling beneficiary households.
CWAC representation in the ACC is important, particularly for intervention communities given the
oversight role of the ACC in managing the CGP and handling grievances. We see strong program impact
on community representation in the ACC, with intervention communities having significantly higher
representation in these committees (93 percent) compared with the control communities (62 percent).
Given the oversight relationship of the ACC to the CWAC, having greater representation facilitates better
communication and management for the CGP.
At the community level, the CGP could also influence the structure and operations of CWACs. The
program improves the frequency that CWACs meet by 59 percentage points, with 96 percent of
intervention CWACs meeting at least quarterly but only 51 percent of control CWACs meeting with the
same frequency. Given that the CGP program relies on the CWACs to support program administration,
81
the regularity of meetings is encouraging. However, there is no program impact on the composition of
the CWACs. The majority (93 percent) of CWACs have an elected executive committee, indicating that
most CWACs are following the appropriate protocols and have the expected structures to function well.
On average, each committee has 9 members, with an average of 3.8 female members per committee, a
similar profile to that at baseline. Similarly, the program does not increase the involvement of women in
leadership. We find that only 16 percent of CWAC chairpersons are female, with no statistical
differences between intervention and control communities, although the program does not affect the
composition of the CWAC.
Table A6.2: Impact of CGP on Community Governance
Program
Impact
Baseline 24-Month
Intervention
24-Month
Control
Community has representation in the local ACC
0.398
(3.11) 0.48 0.93 0.62
Community has an elected executive committee
0.047
(1.19) 0.85 0.98 0.89
Gender of CWAC chairperson is male
0.003
(0.04) 0.94 0.82 0.86
CWAC committee meets regularly (at least
quarterly)
0.587
(4.55) 0.38 0.96 0.49
N 84 45 45
NOTE: Estimations use difference-in-difference modeling in sample communities. T-statistics are in parentheses. Bold
indicates that they are significant at p < .05. All estimations control for district effects and a vector of cluster-level prices.
